{ "id": "AIBD-0042", "title": "Baricitinib, identified with BenevolentAI, reduced COVID-19 deaths in trials", "summary": "Researchers used the Benevolent Platform to identify baricitinib as a candidate treatment for COVID-19.", "date_observed": "2021-09-01", "date_basis": "source publication", "date_precision": "day", "benefit_status": "mixed", "source_collection": null, "source_entry": null, "account": "BenevolentAI researchers reported in February 2020 that their knowledge graph had helped them identify baricitinib, an arthritis drug, as a possible treatment for COVID-19. They proposed clinical testing after reviewing its effects on inflammation and proteins involved in viral entry. In ACTT-2, 1,033 hospitalized adults received remdesivir with either baricitinib or a placebo. Median recovery took 7 days with baricitinib and 8 days with the placebo.\n\nCOV-BARRIER enrolled 1,525 adults. By day 28, 8% of patients receiving baricitinib had died, compared with 13% receiving a placebo. Both groups received standard care. Mortality was a secondary outcome. The trial did not meet its primary outcome of reduced disease progression, so its mortality result was not conclusive under the planned procedure for multiple comparisons.\n\nThe subsequent RECOVERY trial enrolled 8,156 patients. Mortality, its primary outcome, was 12% with baricitinib and 14% with usual care alone.", "category": "benefit created", "unexpected": false, "purpose": "Researchers searched for existing drugs that might treat COVID-19, using a biomedical knowledge graph that combines databases with information extracted from scientific papers.", "system": "Benevolent Platform", "model_version": null, "developer": "BenevolentAI", "deployer": "BenevolentAI research team", "location": "Multicountry trials (ACTT-2, COV-BARRIER, and RECOVERY): Argentina; Brazil; Denmark; Germany; India; Italy; Japan; Mexico; Russia; Singapore; South Korea; Spain; United Kingdom; United States", "domain": "Biomedicine / infectious disease therapeutics", "segment": "Hospitalized COVID-19 patients requiring supplemental oxygen or ventilatory support", "benefit_type": "Improved Physical Health", "taxonomy_note": "The clinical benefit is reduced mortality and faster recovery. AI assisted the researchers who proposed the drug for testing.", "registry_ids": {}, "ratings": [ { "group": "personal", "depth": 5, "reason": "Fewer patients died with baricitinib in COV-BARRIER and RECOVERY. RECOVERY confirmed lower 28-day mortality as its primary outcome in 8,156 patients." }, { "group": "organizational", "depth": 2, "reason": "ACTT-2 reported median recovery of 7 days with baricitinib and 8 days without it. That could free hospital capacity. A separate economic model estimated lower hospital costs, but neither establishes realized staffing or cash savings." }, { "group": "societal", "depth": 4, "reason": "WHO recommended baricitinib with corticosteroids for severe or critical COVID-19. That recommendation supports wider access to a treatment that reduced deaths in trials. Worldwide uptake and population-wide effects were not measured." } ], "ai_role": { "status": "assessed", "statement": "BenevolentAI researchers used the Benevolent Platform to search biomedical relationships and examine networks of proteins. They selected baricitinib after reviewing its targets, dosage, and pharmacology. The clinical investigators then tested the drug. None of these trials compared discovery with AI against discovery without it. The original correspondence and methods paper document AI-assisted selection. FDA authorization and WHO guidance concern the treatment." }, "evidence": [ { "kind": "peer-reviewed study", "citation": "Marconi VC, Ramanan AV, de Bono S, et al. (2021). Efficacy and safety of baricitinib for the treatment of hospitalised adults with COVID-19 (COV-BARRIER). Lancet Respir Med. Results and Figure 2; supplementary appendix 6, pp. 9–14, including the prespecified testing hierarchy.", "title": "Efficacy and safety of baricitinib for the treatment of hospitalised adults with COVID-19 (COV-BARRIER): a randomised, double-blind, parallel-group, placebo-controlled phase 3 trial", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "date": "2021-09-01", "written_by": "evaluator", "quote": "The 28-day all-cause mortality was 8% (n=62) for baricitinib and 13% (n=100) for placebo (hazard ratio [HR] 0·57 [95% CI 0·41–0·78]; nominal p=0·0018)" }, { "kind": "peer-reviewed study", "citation": "RECOVERY Collaborative Group. “Baricitinib in Patients Admitted to Hospital with COVID-19 (RECOVERY): A Randomised, Controlled, Open-Label, Platform Trial and Updated Meta-Analysis.” The Lancet 400 (2022): 359–368. Table 2, Figure 3, and Data sharing. https://doi.org/10.1016/S0140-6736(22)01109-6.", "title": "Baricitinib in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial and updated meta-analysis", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "date": "2022-07-28", "written_by": "independent", "quote": null }, { "kind": "other", "citation": "Richardson P, Griffin I, Tucker C, Smith D, Oechsle O, Phelan A, Rawling M, Savory E, Stebbing J (2020). Baricitinib as potential treatment for 2019-nCoV acute respiratory disease. Lancet.", "title": "Baricitinib as potential treatment for 2019-nCoV acute respiratory disease", "url": "https://europepmc.org/article/PMC/PMC7137985", "date": "2020-02-04", "written_by": "makers", "quote": "BenevolentAI's knowledge graph is a large repository of structured medical information, including numerous connections extracted from scientific literature by machine learning." }, { "kind": "peer-reviewed study", "citation": "Daniel P. Smith et al. “Expert-Augmented Computational Drug Repurposing Identified Baricitinib as a Treatment for COVID-19.” Frontiers in Pharmacology 12 (July 28, 2021): 709856. https://doi.org/10.3389/fphar.2021.709856.", "title": "Expert-Augmented Computational Drug Repurposing Identified Baricitinib as a Treatment for COVID-19", "url": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "date": "2021-07-28", "written_by": "makers", "quote": null }, { "kind": "peer-reviewed study", "citation": "Kalil AC, Patterson TF, Mehta AK, et al. (2021). Baricitinib plus Remdesivir for Hospitalized Adults with Covid-19. N Engl J Med. Table 2 and trial oversight; supplementary appendix, pp. 19–23; protocol NCT04401579.", "title": "Baricitinib plus Remdesivir for Hospitalized Adults with Covid-19", "url": "https://europepmc.org/article/PMC/PMC7745180", "date": "2020-12-11", "written_by": "independent", "quote": "median time to recovery of 7 days (95% confidence interval [CI], 6 to 8), as compared with 8 days (95% CI, 7 to 9)" }, { "kind": "press release", "citation": "U.S. Food and Drug Administration. “Coronavirus (COVID-19) Update: FDA Authorizes Drug Combination for Treatment of COVID-19.” November 19, 2020. Authorization, evidentiary standard, and ACTT-2 results.", "title": "FDA Authorizes Drug Combination for Treatment of COVID-19", "url": "https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-authorizes-drug-combination-treatment-covid-19", "date": "2020-11-19", "written_by": "independent", "quote": "the drug baricitinib, in combination with remdesivir, for the treatment of suspected or laboratory confirmed COVID-19" }, { "kind": "report", "citation": "World Health Organization (2022). WHO recommends two new drugs to treat COVID-19.", "title": "WHO recommends two new drugs to treat COVID-19", "url": "https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "date": "2022-01-14", "written_by": "independent", "quote": "The first drug, baricitinib, is strongly recommended for patients with severe or critical COVID-19." }, { "kind": "report", "citation": "BenevolentAI. “BenevolentAI Unaudited Preliminary Results for the Year Ended 31 December 2022,” March 16, 2023, Operational highlights, baricitinib bullet.", "title": "BenevolentAI Unaudited Preliminary Results for the Year Ended 31 December 2022", "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "date": "2023-03-16", "written_by": "makers", "quote": null }, { "kind": "press release", "citation": "BenevolentAI. “BenevolentAI’s platform-derived hypothesis for COVID-19 treatment validated in US NIAID randomised control trial.” September 14, 2020. “About BenevolentAI,” headquarters statement.", "title": "BenevolentAI’s platform-derived hypothesis for COVID-19 treatment validated in US NIAID randomised control trial", "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "date": "2020-09-14", "written_by": "makers", "quote": null }, { "kind": "peer-reviewed study", "citation": "Robert Ohsfeldt et al. “Cost-Effectiveness of Baricitinib Compared With Standard of Care: A Modeling Study in Hospitalized Patients With COVID-19 in the United States.” Clinical Therapeutics 43, no. 11 (2021): 1877–1893.e4. Tables IV–V and supplemental table S6. https://doi.org/10.1016/j.clinthera.2021.09.016.", "title": "Cost-Effectiveness of Baricitinib Compared With Standard of Care: A Modeling Study in Hospitalized Patients With COVID-19 in the United States", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8487786/", "date": "2021-10-04", "written_by": "evaluator", "quote": null } ], "replication": { "produced_by": [ "BenevolentAI researchers combined a biomedical knowledge graph with expert review to select baricitinib.", "Clinical investigators separately tested the drug alongside the care available in 2020 and 2021." ], "must_hold": [ "The clinical comparisons concern hospitalized patients and the eligibility criteria of each trial.", "Background treatment differed: remdesivir in ACTT-2, and mostly corticosteroids in COV-BARRIER and RECOVERY.", "Reproducing the AI selection would also require the January 2020 graph, software configuration, and researchers’ decisions. The complete snapshot is not public." ], "echoes": [ { "date": "2021-09-01", "setting": "COV-BARRIER: 1,525 adults. Lower mortality was a secondary finding after an unsuccessful primary comparison.", "model_version": null, "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/" }, { "date": "2022-07-28", "setting": "RECOVERY: 8,156 patients in 159 U.K. hospitals. Lower 28-day mortality was the primary finding. These were further tests of the drug, not repetitions of the AI discovery process.", "model_version": null, "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/" } ] }, "paired_incident": null, "headline": "An arthritis drug helped patients with severe COVID-19", "metric": "28-day all-cause mortality, a secondary endpoint, was 8% versus 13%", "comparator": { "against": "ACTT-2: remdesivir plus placebo versus remdesivir plus baricitinib", "before": "8 days (95% CI, 7 to 9)", "after": "7 days (95% confidence interval [CI], 6 to 8)", "basis": "measured", "quote": "Patients receiving baricitinib had a median time to recovery of 7 days (95% confidence interval [CI], 6 to 8), as compared with 8 days (95% CI, 7 to 9)", "url": "https://europepmc.org/article/PMC/PMC7745180" }, "later": { "status": "persisted", "date": "2022-07-28", "note": "RECOVERY independently found lower 28-day mortality with baricitinib. WHO had recommended its use with corticosteroids for severe or critical COVID-19 in January 2022. The recommendation does not establish worldwide uptake.", "quote": null, "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/" }, "value": null, "developer_country": "United Kingdom", "deployer_country": "United Kingdom", "unrated_because": null, "brief_notes": null, "brief": null, "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "source": null, "editor": "AIBD" }, { "date": "2026-10-20", "change": "Paired with AI Incident Database incident 225 (medium confidence); severity not given by the source.", "source": "https://incidentdatabase.ai/cite/225", "editor": "AIBD" }, { "date": "2026-10-20", "change": "Recategorised from specific harm prevented to 1 because the sources show an AI-generated repurposing hypothesis confirmed at cohort level by randomized trials rather than a specific harm prevented for an identifiable beneficiary, and, on the fact-checker's other findings, made the ACTT-2 quote verbatim, relabelled the Lancet correspondence letter as “other”, used the exact source titles, replaced the unsupported “hospitals worldwide” with the trial sponsors as deployers, and stated in the summary and account that COV-BARRIER missed its primary endpoint and that its 28-day mortality reduction was a secondary endpoint.", "source": "Richardson P, et al. Lancet 2020 (https://europepmc.org/article/PMC/PMC7137985); Kalil AC, et al. N Engl J Med 2021 (https://europepmc.org/article/PMC/PMC7745180); Marconi VC, et al. Lancet Respir Med 2021 (https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/)", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Added what the benefit is measured against, what happened since from the cited sources.", "source": "https://europepmc.org/article/PMC/PMC7745180; https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Improved Physical Health under Trustible's full taxonomy of 23 types; it was Improved Knowledge Discovery among the six first used.", "source": "https://trustible.ai/post/a-framework-for-measuring-the-benefits-of-ai/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Country of the developer and the deployer added from the organization's own site or a registry.", "source": "https://www.benevolent.com/contact-us/; https://en.wikipedia.org/wiki/National_Institute_of_Allergy_and_Infectious_Diseases", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the replication conditions rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "source": null, "editor": "AIBD" }, { "date": "2026-10-23", "change": "Personal benefit depth changed from 4 to 5 for the reported life-preserving outcome.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Traced the AI hypothesis, recovery and mortality findings, and regulatory actions to their original sources. Put COV-BARRIER first for the mortality claim, preserved the failed primary endpoint and added the multiplicity limitation. Removed the generic FDA drug-list page from active evidence because it adds no case-specific support beyond the retained authorization. Distinguished NIAID-led ACTT-2 from the Lilly-sponsored COV-BARRIER analysis.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Rewrote the short title and summary in plain language, preserving the reported benefit and material limits. Added the named publication or institution to the reviewed lead source for homepage display.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Identified the AI system or feature in the supporting summary, or stated that the source does not name the software. Preserved the reported benefit and material qualifications.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Traced the system identity beyond the original case report. Added product names and supporting original sources, preserving historical-version limits and any inference in the source notes.", "source": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Shortened the homepage supporting statement to one sentence naming the system and its contribution; retained detailed outcomes and qualifications in the case account and source notes.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Revised the title and subtitle for natural phrasing, with named researchers, shorter dependencies, and less compressed terminology; preserved the system identities and evidence qualifications.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Added the developers’ original methods paper, identified the SVO component and January 2020 workflow, and clarified the scientists’ role in selecting baricitinib. The exact model release remains unreported.", "source": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Revised the homepage subtitle for concise database prose, with courtesy titles for named beneficiaries and numbered clauses for the Coates drug-selection account. Preserved the reported evidence and material uncertainty.", "source": null, "editor": "AIBD" }, { "date": "2026-10-25", "change": "Corrected the developer jurisdiction to its contemporaneous U.K. headquarters and replaced the London-only location with the 14 countries in ACTT-2 and COV-BARRIER. Changed the chronology from the FDA authorization date to September 1, 2021, the original publication date of the COV-BARRIER mortality findings. Trial countries do not imply a separately demonstrated benefit in every country.", "source": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "editor": "AIBD" }, { "date": "2026-10-25", "change": "Rewrote the account and resolved all mandatory fields. Added RECOVERY and the original economic model, corrected the AI operator, distinguished projected wider benefits from observed patient outcomes, replaced WHO guidance counted as replication, and removed an unrelated Watson incident. Missing model and attribution details remain explicit. Reordered the sources by their contribution and remapped their identifiers.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "editor": "AIBD, Codex-assisted review" }, { "date": "2026-10-25", "change": "Added historical control from original corporate filings, documented funding and scientific-resource relationships, a labeled strategic assessment, and derived economic scenarios. Preserved the unreported exact model version, with a documented configuration proxy. Distinguished reported findings, calculations, and inferences throughout the assessment.", "source": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "editor": "AIBD, Codex-assisted review" } ], "confidence": "studied", "shade": "high", "claim_status": "Independently studied", "claim_basis": "Sources: two independent peer-reviewed studies, one independent report, one independent press release, the makers' peer-reviewed study, the makers' report, the makers' press release, the makers' other and two peer-reviewed studies from the evaluators. The cited evidence includes no finding by a regulator or court.", "profile": { "case_id": "AIBD-0042", "catalog_version": "1.0.0", "source_fingerprint": "a8e1b0052ad9c7eff39b20f5389ce8ba29abb827c2271690fcac5264ab7db26f", "collection_refs": [ { "collection_id": "aibd", "version": "snapshot-8bf7cfec9cb185aa" }, { "collection_id": "aibd", "version": "2026-10-25" }, { "collection_id": "querist", "version": "2026-10-25" }, { "collection_id": "viator", "version": "2026-10-25" } ], "source_ids": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010" ], "families": { "strategic assessments": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "AIBD-0042-strategic-uk-preparedness" ], "role": "strategic assessments" }, "provenance": { "kind": "derivation", "basis": "AIBD has supplied a substantive qualitative assessment with a named perspective, rationale, uncertainty, and sources.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "record_ids": [ "AIBD-0042-strategic-uk-preparedness" ] }, "evaluations": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "ACTT-2", "COV-BARRIER", "RECOVERY-baricitinib", "Ohsfeldt-2021-economic-model" ], "role": "evaluations" }, "provenance": { "kind": "source_review", "basis": "The three original trial reports and economic-model publication identify distinct evaluations.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "record_ids": [ "ACTT-2", "COV-BARRIER", "RECOVERY-baricitinib", "Ohsfeldt-2021-economic-model" ] }, "monetary observations": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "Ohsfeldt-2021-hospital-model", "Ohsfeldt-2021-payer-model" ], "role": "monetary observations" }, "provenance": { "kind": "source_review", "basis": "The original study reports separate hospital and payer model results. These are treatment valuations, not attributed AI savings.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "record_ids": [ "Ohsfeldt-2021-hospital-model", "Ohsfeldt-2021-payer-model" ] }, "institutional decisions": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "FDA-baricitinib-EUA-2020-11-19", "WHO-baricitinib-guidance-2022-01-14" ], "role": "institutional decisions" }, "provenance": { "kind": "source_review", "basis": "Direct FDA and WHO statements establish the dated institutional decisions.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "record_ids": [ "FDA-baricitinib-EUA-2020-11-19", "WHO-baricitinib-guidance-2022-01-14" ] } }, "fields": { "actors[]": { "state": "value_present", "value": { "kind": "structured", "data": { "actors": [ { "name": "BenevolentAI", "roles": [ "AI developer", "AI provider", "AI operator" ] }, { "name": "BenevolentAI research team", "roles": [ "Graph analysis", "Drug selection" ] }, { "name": "NIAID and the ACTT-2 Study Group", "roles": [ "Clinical investigators", "ACTT-2 sponsor" ] }, { "name": "Eli Lilly and Company", "roles": [ "Baricitinib supplier", "COV-BARRIER sponsor", "Economic-model funder" ] }, { "name": "COV-BARRIER investigators", "roles": [ "Clinical investigators" ] }, { "name": "University of Oxford and the RECOVERY Collaborative Group", "roles": [ "Trial sponsor", "Clinical investigators" ] }, { "name": "Gilead Sciences", "roles": [ "Remdesivir supplier for ACTT-2" ] }, { "name": "U.S. Food and Drug Administration", "roles": [ "Regulator" ] }, { "name": "World Health Organization", "roles": [ "Guideline body" ] }, { "name": "Hospitalized patients with COVID-19", "roles": [ "Treatment recipients", "Clinical beneficiaries" ] }, { "name": "Medical Decision Modeling Inc and the economic-study authors", "roles": [ "Economic modelers" ] } ] } }, "provenance": { "kind": "source_review", "basis": "Separated the AI operator from clinical investigators, suppliers, and policy bodies.", "record_fields": [ "developer", "deployer" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "BenevolentAI researchers selected the drug. U.S. government researchers, Lilly, and Oxford organized trials. The FDA authorized treatment, and WHO recommended it." } }, "actors[].ownership_control": { "state": "value_present", "value": { "kind": "structured", "data": { "reported_status": "Historical control is documented in the company’s 2019 accounts.", "entity": "BenevolentAI Limited", "company_number": "09781806", "jurisdiction": "United Kingdom", "as_of": "2019-12-31", "controlling_person": "Kenneth Mulvany, a director and shareholder", "source_basis": "The accounts identify Mulvany as the person who controlled the company.", "aibd_calculated_voting_share": { "person": "Kenneth Patrick Mulvany", "snapshot_date": "2019-10-15", "ordinary_shares_held": 881000, "ordinary_shares_issued": 1831829, "preferred_shares_issued": 208623, "formula": "881,000 ÷ (1,831,829 + 208,623) × 100", "percent": 43.18, "interpretation": "Share of general voting shares in the October register. This is not a measure of all economic equity or proof of majority ownership.", "capital_check": "The December 2019 allotment left the ordinary and preferred counts unchanged. Its new G2 growth shares did not carry ordinary voting rights before an exit, apart from class resolutions.", "confidence": "High for the calculation from the filed snapshot. The snapshot does not establish every beneficial holding on an unobserved January 2020 date." }, "investor_context": { "investor": "Temasek", "jurisdiction": "Singapore", "investment_year": 2019, "reported_amount": "$90 million", "amount_note": "The announcements use the dollar sign without spelling out a currency. The amount is preserved as reported and is not converted or assigned to this research.", "government_relationship": "Temasek’s 2019 report identifies Singapore’s Minister for Finance as its sole shareholder. It also states that Temasek does not direct its portfolio companies’ business decisions.", "control_boundary": "An investment from a state-owned institution does not establish state control of BenevolentAI or direction of the baricitinib research." }, "temporal_scope": "Ownership surrounding the January 2020 discovery, not the developer’s present corporate structure." } }, "provenance": { "kind": "source_review", "basis": "Read original corporate filings and transaction announcements. The controlling person is reported directly; the voting-share percentage is a separate AIBD calculation.", "record_fields": [], "source_refs": [ "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Kenneth Mulvany controlled BenevolentAI at the end of 2019 and held 43.18% of its general voting shares in October.", "editor_note": "The percentage is calculated from his 881,000 shares and the 2,040,452 voting shares in the 15 October 2019 register (reference 12). The December allotment left that total unchanged, but individual holdings could have changed before the January research." } }, "ai_contribution": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ai_role": { "status": "assessed", "statement": "BenevolentAI researchers used the Benevolent Platform to search biomedical relationships and examine networks of proteins. They selected baricitinib after reviewing its targets, dosage, and pharmacology. The clinical investigators then tested the drug. None of these trials compared discovery with AI against discovery without it. The original correspondence and methods paper document AI-assisted selection. FDA authorization and WHO guidance concern the treatment." } } }, "provenance": { "kind": "source_review", "basis": "The authors document AI-assisted candidate selection; no controlled comparison isolates its incremental contribution.", "record_fields": [ "ai_role" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used the system to find relationships among proteins, diseases, and medicines before selecting baricitinib." } }, "ai_relationship": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "category": "benefit created", "ai_role": { "status": "assessed", "statement": "BenevolentAI researchers used the Benevolent Platform to search biomedical relationships and examine networks of proteins. They selected baricitinib after reviewing its targets, dosage, and pharmacology. The clinical investigators then tested the drug. None of these trials compared discovery with AI against discovery without it. The original correspondence and methods paper document AI-assisted selection. FDA authorization and WHO guidance concern the treatment." } } }, "provenance": { "kind": "source_review", "basis": "Direct use of AI-assisted discovery preceded separate clinical testing of the proposed drug.", "record_fields": [ "category", "ai_role" ], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used AI to investigate a new use for an existing medicine." } }, "assessments[].alternatives": { "state": "value_present", "value": { "kind": "structured", "data": { "personal": "Depth 5 is supported by mortality outcomes. The AI contribution is less certain than the drug benefit.", "organizational": "If only measured organizational gains were admitted, this rating would remain unresolved. The retained depth describes the potential capacity benefit.", "societal": "Depth 4 describes intended wider harm prevention. A recommendation alone does not establish a realized population-wide change." } }, "provenance": { "kind": "source_review", "basis": "Preserved the consequential alternatives rather than averaging different kinds of benefit.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The hospital and societal grades describe possible benefits. A rule requiring observed benefits would leave those two grades unassigned." } }, "assessments[].assessor_date": { "state": "value_present", "value": { "kind": "structured", "data": { "assessor": "Codex-assisted AIBD review", "reviewed_on": "2026-10-25", "human_approval": "Not represented as an independent human review." } }, "provenance": { "kind": "curatorial", "basis": "Recorded the actual review method without attributing unperformed verification to Peter.", "record_fields": [], "source_refs": [], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "AIBD, assisted by Codex, on 25 October 2026." } }, "assessments[].depth": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "personal", "depth": 5, "reason": "Fewer patients died with baricitinib in COV-BARRIER and RECOVERY. RECOVERY confirmed lower 28-day mortality as its primary outcome in 8,156 patients." }, { "group": "organizational", "depth": 2, "reason": "ACTT-2 reported median recovery of 7 days with baricitinib and 8 days without it. That could free hospital capacity. A separate economic model estimated lower hospital costs, but neither establishes realized staffing or cash savings." }, { "group": "societal", "depth": 4, "reason": "WHO recommended baricitinib with corticosteroids for severe or critical COVID-19. That recommendation supports wider access to a treatment that reduced deaths in trials. Worldwide uptake and population-wide effects were not measured." } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed each stakeholder-specific rationale. Depth describes the stated benefit; support and realization are separate.", "record_fields": [ "ratings" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "For patients: 5 for fewer deaths and 2 for faster recovery. The broader assessments remain 2 for organizations and 4 for society." } }, "assessments[].display_confidence": { "state": "value_present", "value": { "kind": "structured", "data": { "confidence": "high", "evidence_word": "studied", "claim_status": "Independently studied", "basis": "Drug effects are supported by randomized trials, including an independent trial. The developers document the AI-assisted selection, but its incremental clinical contribution remains unmeasured.", "interpretation": "The shared solid fill describes the clinical evidence. It does not establish that all organizational or societal effects were measured." } }, "provenance": { "kind": "source_review", "basis": "Recomputed the display from the corrected source types and preserved the distinction between treatment and discovery evidence.", "record_fields": [ "evidence" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The solid fill denotes independently studied treatment benefits. It does not mean that the projected hospital and societal benefits occurred." } }, "assessments[].rationale": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "personal", "depth": 5, "reason": "Fewer patients died with baricitinib in COV-BARRIER and RECOVERY. RECOVERY confirmed lower 28-day mortality as its primary outcome in 8,156 patients." }, { "group": "organizational", "depth": 2, "reason": "ACTT-2 reported median recovery of 7 days with baricitinib and 8 days without it. That could free hospital capacity. A separate economic model estimated lower hospital costs, but neither establishes realized staffing or cash savings." }, { "group": "societal", "depth": 4, "reason": "WHO recommended baricitinib with corticosteroids for severe or critical COVID-19. That recommendation supports wider access to a treatment that reduced deaths in trials. Worldwide uptake and population-wide effects were not measured." } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed each stakeholder-specific rationale. Depth describes the stated benefit; support and realization are separate.", "record_fields": [ "ratings" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Lower mortality preserves lives. Faster recovery is a measurable health improvement. Shorter hospital stays could free capacity, and wider access could extend the treatment’s benefits." } }, "assessments[].rubric_version": { "state": "value_present", "value": { "kind": "structured", "data": { "rubric": "AIBD five-level depth scale and three beneficiary groups", "source_file": "aibd/templates/method.html", "source_revision": "4e31895", "applied_on": "2026-10-25", "personal_5": "Life preserved", "organizational_2": "Improved operational capability", "societal_4": "Serious harm averted; wider benefit remains intended rather than measured" } }, "provenance": { "kind": "curatorial", "basis": "Recorded the local rubric used for this review.", "record_fields": [], "source_refs": [], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "AIBD’s five-level scale, applied on 25 October 2026." } }, "assessments[].status_basis": { "state": "value_present", "value": { "kind": "structured", "data": { "personal": "Realized clinical benefit in randomized trials", "organizational": "Potential capacity improvement and modeled costs", "societal": "Observed recommendation, intended wider health benefit" } }, "provenance": { "kind": "source_review", "basis": "Separated realization from the depth of the benefit.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Trials established a benefit for patients. Hospital savings were modeled, and wider health benefits remain intended." } }, "assessments[].target": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "personal", "depth": 5, "reason": "Fewer patients died with baricitinib in COV-BARRIER and RECOVERY. RECOVERY confirmed lower 28-day mortality as its primary outcome in 8,156 patients." }, { "group": "organizational", "depth": 2, "reason": "ACTT-2 reported median recovery of 7 days with baricitinib and 8 days without it. That could free hospital capacity. A separate economic model estimated lower hospital costs, but neither establishes realized staffing or cash savings." }, { "group": "societal", "depth": 4, "reason": "WHO recommended baricitinib with corticosteroids for severe or critical COVID-19. That recommendation supports wider access to a treatment that reduced deaths in trials. Worldwide uptake and population-wide effects were not measured." } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed each stakeholder-specific rationale. Depth describes the stated benefit; support and realization are separate.", "record_fields": [ "ratings" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Hospitalized patients, hospitals, and the wider population that could receive treatment." } }, "availability": { "state": "value_present", "value": { "kind": "structured", "data": { "historical_ai_access": "BenevolentAI scientists used an internal research platform. The reviewed methods paper does not provide public access to the complete 2020 system.", "public_materials": "Published workflow, source databases, clinical articles, and linked trial materials.", "procurement_price": "Not reported for the discovery system.", "current_access": "Current licensing, geographic availability, and access to the historical snapshot are unresolved.", "treatment": "Baricitinib is a separate prescription medicine; the cited FDA and WHO statements describe historical authorization and guidance, not current individualized treatment advice.", "reviewed_on": "2026-10-25" } }, "provenance": { "kind": "source_review", "basis": "Separated access to the AI workflow from access to the drug.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "BenevolentAI used its internal platform. Its methods are published, but the complete January 2020 system is unavailable. Baricitinib is a prescription medicine." } }, "brief": { "state": "value_present", "value": { "kind": "text", "text": "Researchers used the Benevolent Platform to identify baricitinib for COVID-19 testing. Randomized trials later found faster recovery or fewer deaths among hospitalized patients. The trials tested the drug rather than the additional value of AI-assisted discovery." }, "provenance": { "kind": "source_review", "basis": "The summary draws on the original discovery and clinical publications.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used AI to identify baricitinib for COVID-19 trials. Clinical investigators later found that patients recovered sooner and fewer died." } }, "case_kind": { "state": "value_present", "value": { "kind": "structured", "data": { "unit": "AI-assisted drug repurposing with subsequent clinical evaluations", "counting": "One discovery episode. The trials and articles are linked evidence, not additional AIBD cases." } }, "provenance": { "kind": "source_review", "basis": "Assigned the unit to the documented discovery episode and its follow-up.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Repurposing a medicine for COVID-19." } }, "case_links[]": { "state": "value_present", "value": { "kind": "structured", "data": { "related_evaluations": [ "NCT04401579", "NCT04421027", "NCT04381936" ], "same_discovery": "The three trials concern the drug selected in the January 2020 BenevolentAI work.", "duplicate_search": "No other existing AIBD record mentions baricitinib." } }, "provenance": { "kind": "curatorial", "basis": "Reviewed the local collection for duplicates and linked the original trial registrations.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "ACTT-2, COV-BARRIER, and RECOVERY tested baricitinib after the initial proposal." } }, "claims[].evidence_links": { "state": "value_present", "value": { "kind": "structured", "data": { "links": [ { "claim": "AIBD-0042-claim-discovery", "support": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "locator": "Original correspondence; Methods and Results of the methods paper" }, { "claim": "AIBD-0042-claim-recovery", "support": [ "AIBD-0042-source-0005" ], "locator": "Table 2" }, { "claim": "AIBD-0042-claim-mortality", "support": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002" ], "locator": "COV-BARRIER Figure 2 and statistical hierarchy; RECOVERY Table 2" }, { "claim": "AIBD-0042-claim-policy", "support": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007" ], "locator": "Dated agency statements" }, { "claim": "AIBD-0042-claim-economics", "support": [ "AIBD-0042-source-0010" ], "locator": "Tables IV–V and supplemental table S6" } ], "qualifications": [ { "claim": "AIBD-0042-claim-mortality", "source": "AIBD-0042-source-0001", "qualification": "The COV-BARRIER primary outcome was not statistically significant; mortality was secondary." }, { "claim": "AIBD-0042-claim-economics", "source": "AIBD-0042-source-0010", "qualification": "Hospital savings reverse to higher costs in the mortality-only scenario." } ] } }, "provenance": { "kind": "source_review", "basis": "Mapped the statements to original findings and their material qualifications.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "References 3–4 describe the discovery, 5 reports recovery, 1–2 report mortality, 6–7 record agency decisions, and 10 estimates costs." } }, "claims[].evidence_strength": { "state": "value_present", "value": { "kind": "structured", "data": { "clinical_outcomes": "Randomized trials support a drug benefit; RECOVERY independently confirms lower mortality.", "ai_use": "Original correspondence and detailed developer methods establish the reported workflow.", "ai_increment": "No controlled estimate of the benefit added by using AI.", "institutional_actions": "Direct agency statements establish the decisions and dates.", "wider_impact": "Hospital capacity and population-wide benefits were not measured. Economic effects are modeled." } }, "provenance": { "kind": "source_review", "basis": "Assessed each causal link separately.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Randomized trials establish the treatment benefit. The developers describe the AI workflow. No trial compares discovery with and without AI." } }, "claims[].missingness": { "state": "value_present", "value": { "kind": "structured", "data": { "model_release": "Not reported", "frozen_discovery_snapshot": "Not reported", "incremental_ai_benefit": "No controlled numerical uplift is reported. AIBD supplies a qualitative inference and documented workflow proxies.", "realized_hospital_savings": "Not reported", "worldwide_beneficiary_count": "Not reported", "control_of_developer_in_2020": "Historical control is established by the 2019 accounts. The separate voting-share calculation uses the October 2019 snapshot.", "cross_border_transfers": "Company funding is documented. Scientific-resource reuse is documented, with geographic links labeled as inferences. No patient-data route or case-level funding allocation is reported.", "inaccessible_upstream_materials": "See source access notes for the Sorrell supplement and the original WHO guideline PDF.", "return_payback": "No source-reported ROI or payment timetable. AIBD supplies labeled value scenarios and a reference ratio.", "strategic_effect": "AIBD provides a qualitative preparedness assessment. No quantitative national advantage is reported or invented." } }, "provenance": { "kind": "source_review", "basis": "Recorded consequential gaps without substituting zero or assuming that missing evidence disproves benefit.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The publications omit the exact software release, a comparison without AI, observed savings, and the number of people helped worldwide." } }, "claims[].proposition": { "state": "value_present", "value": { "kind": "structured", "data": { "claims": [ { "id": "AIBD-0042-claim-discovery", "statement": "BenevolentAI researchers used their graph to identify baricitinib for COVID-19 testing.", "type": "Original researchers’ account", "sources": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "locator": "Original correspondence; Methods and Results of the methods paper" }, { "id": "AIBD-0042-claim-recovery", "statement": "ACTT-2 found faster recovery with baricitinib added to remdesivir.", "type": "Randomized clinical finding", "sources": [ "AIBD-0042-source-0005" ], "locator": "Table 2" }, { "id": "AIBD-0042-claim-mortality", "statement": "COV-BARRIER and RECOVERY found lower mortality with baricitinib.", "type": "Randomized clinical findings with different primary outcomes", "sources": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002" ], "locator": "COV-BARRIER Figure 2 and statistical hierarchy; RECOVERY Table 2" }, { "id": "AIBD-0042-claim-policy", "statement": "FDA authorized use in November 2020 and WHO recommended use in January 2022.", "type": "Documented institutional decisions", "sources": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007" ], "locator": "Dated agency statements" }, { "id": "AIBD-0042-claim-economics", "statement": "A published model estimated treatment costs and health gains.", "type": "Sponsor-funded model, not observed AI savings", "sources": [ "AIBD-0042-source-0010" ], "locator": "Tables IV–V and supplemental table S6" } ] } }, "provenance": { "kind": "source_review", "basis": "Separated discovery, clinical, institutional, and modeled economic statements.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used AI to select baricitinib for testing. Clinical investigators subsequently found that patients recovered faster and fewer died." } }, "cross_border_flows[]": { "state": "value_present", "value": { "kind": "structured", "data": { "reported_status": "The funding relationship and reuse of named scientific resources are documented. Geographic knowledge flows are inferred from the resource producers and the U.K. discovery team.", "flows": [ { "type": "Investment", "origin": "Singapore", "destination": "United Kingdom", "from_actor": "Temasek", "to_actor": "BenevolentAI", "year": 2019, "reported_amount": "$90 million", "evidence_status": "Reported by the recipient and its transaction counsel", "scope": "Company-wide development of the platform and drug programs. The sources do not identify a baricitinib budget or the bank settlement route.", "source_refs": [ "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek" ] }, { "type": "Reuse of scientific information", "origin": "Japan", "destination": "United Kingdom", "resource": "KEGG", "evidence_status": "Documented use; geographic relationship inferred", "basis": "The methods paper names KEGG. Its developers trace the resource to the Japanese Human Genome Project.", "source_refs": [ "AIBD-0042-source-0004", "https://www.genome.jp/kegg/kegg1.html" ] }, { "type": "Reuse of scientific information", "origin": "Italy", "destination": "United Kingdom", "resource": "SIGNOR", "evidence_status": "Documented use; geographic relationship inferred", "basis": "The methods paper names SIGNOR. Its resource paper identifies an Italy-led collaboration that also includes U.K. researchers.", "source_refs": [ "AIBD-0042-source-0004", "https://doi.org/10.1093/nar/gkz949" ] }, { "type": "Published research followed by clinical testing", "origin": "United Kingdom", "destination": "U.S.-led ACTT-2 and the countries in the multinational COV-BARRIER trial", "evidence_status": "Documented sequence in the discovery and trial publications", "scope": "The treatment hypothesis crossed research settings. This does not mean that the original AI system was deployed at the trial sites.", "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0001" ] } ], "boundary": "Resource origins do not establish server locations, network routes, or transfers of patient data. The funding amount belongs to the company, not to this individual benefit.", "confidence": "High for the named funding relationship and resource use. Moderate for the inferred geographic relationships." } }, "provenance": { "kind": "source_review", "basis": "Combined direct transaction evidence, original resource histories, and the discovery methods. Kept documented relationships separate from inferred geographic links.", "record_fields": [], "source_refs": [ "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "AIBD-0042-source-0004", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "AIBD-0042-source-0003", "AIBD-0042-source-0005", "AIBD-0042-source-0001" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Singapore’s Temasek invested in BenevolentAI. The U.K. researchers used Japanese and Italian scientific resources before trials in several countries.", "editor_note": "The geographic links follow the producers of KEGG and SIGNOR, which Smith and colleagues name in reference 4. They identify international knowledge reuse, not the route of a data transfer. Temasek funded the company as a whole." } }, "decisions[].action_outcome": { "state": "value_present", "value": { "kind": "structured", "data": { "FDA": { "date": "2020-11-19", "action": "Issued the EUA to Eli Lilly for baricitinib with remdesivir in eligible hospitalized adults and children aged two or older requiring specified respiratory support." }, "WHO": { "date": "2022-01-14", "action": "Strongly recommended baricitinib with corticosteroids for severe or critical COVID-19." }, "later_evidence": "RECOVERY later confirmed lower 28-day mortality. These sources do not quantify adoption caused by the decisions." } }, "provenance": { "kind": "source_review", "basis": "Separated the decisions, later clinical evidence, and unmeasured implementation effects.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The FDA authorized baricitinib with remdesivir on 19 November 2020. WHO recommended it with corticosteroids on 14 January 2022." } }, "decisions[].grounds": { "state": "value_present", "value": { "kind": "structured", "data": { "FDA": "ACTT-2 results and the emergency-use standard: reasonable belief that the combination may be effective, with known and potential benefits outweighing known and potential risks for the authorized population. An EUA differs from full approval.", "WHO": "Guideline review of clinical-trial evidence. WHO recommended baricitinib with corticosteroids and identified access and affordability as conditions for wider benefit.", "AI_scope": "Neither decision assessed the incremental value of the AI discovery method." } }, "provenance": { "kind": "source_review", "basis": "The FDA announcement states its evidentiary standard; WHO states the recommendation and access limitations.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The FDA relied on ACTT-2. WHO weighed the clinical evidence, harms, affordability, and access." } }, "decisions[].institution_task": { "state": "value_present", "value": { "kind": "structured", "data": { "decisions": [ { "id": "FDA-baricitinib-EUA-2020-11-19", "institution": "U.S. Food and Drug Administration", "jurisdiction": "United States", "task": "Whether to authorize emergency use of baricitinib with remdesivir for specified hospitalized patients with COVID-19." }, { "id": "WHO-baricitinib-guidance-2022-01-14", "institution": "World Health Organization", "scope": "International clinical guidance", "task": "Whether to recommend baricitinib for severe or critical COVID-19." } ] } }, "provenance": { "kind": "source_review", "basis": "Read the original dated agency announcements.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The FDA considered emergency authorization. WHO considered treatment guidance for severe or critical COVID-19." } }, "decisions[].success_criterion": { "state": "value_present", "value": { "kind": "structured", "data": { "clinical_question": "Whether adding baricitinib improves recovery or survival with acceptable harms compared with the same background care.", "adoption_question": "Whether eligible patients can obtain treatment and whether clinical benefits persist in routine practice.", "unresolved": "The cited agency statements do not specify a numerical global uptake target or measure the effects caused by adopting their guidance." } }, "provenance": { "kind": "source_review", "basis": "The clinical questions follow the source outcomes; implementation remains a separate unanswered question.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Both agencies sought better survival or recovery with acceptable harms. They did not quantify the additional use or health improvement that their decisions caused." } }, "deployment_stage": { "state": "value_present", "value": { "kind": "structured", "data": { "ai": "Used by BenevolentAI researchers in January 2020", "treatment": "Evaluated in randomized clinical trials and subsequently authorized or recommended", "scope": "The AI application and the drug have different deployment stages." } }, "provenance": { "kind": "source_review", "basis": "The methods paper documents actual use of the graph; clinical and policy sources concern the resulting treatment.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used the AI system in January 2020. Clinical trials and treatment authorization followed." } }, "economics[].amount": { "state": "value_present", "value": { "kind": "structured", "data": { "hospital_base_case": { "incremental_expenditure_usd": -2436, "incremental_reimbursement_usd": -503, "incremental_net_cost_usd": -1932 }, "hospital_mortality_only": { "incremental_expenditure_usd": 1978, "incremental_reimbursement_usd": 0, "incremental_net_cost_usd": 1978 }, "payer_base_case": { "placebo_soc_total_usd": 329268, "baricitinib_soc_total_usd": 346544, "incremental_total_usd": 17276, "incremental_qalys": 0.6703, "usd_per_qaly_gained": 25774 }, "payer_mortality_only": { "incremental_total_usd": 17902, "incremental_qalys": 0.6664, "usd_per_qaly_gained": 26862 }, "rounding": "The published rounded components do not subtract exactly to the reported net amount of $1,932." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Table IV; appendix Tables S5–S6.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The model estimates US$1,932 less in net hospital costs per patient. Over a lifetime, costs rise US$17,276 and health improves by 0.6703 quality-adjusted life-years." } }, "economics[].costs[]": { "state": "value_present", "value": { "kind": "structured", "data": { "included": [ "Drug acquisition", "Medical care during admission", "Hospital reimbursement for inpatient care", "Postacute care and lifetime healthcare in the payer model", "Missed work in the payer model" ], "unit_inputs": { "baricitinib_2mg_tablet_usd": 75.5, "baricitinib_assumed_dose": "4 mg daily for up to 14 days or discharge", "hospital_day_mechanical_ventilation_usd": 3660, "hospital_day_noninvasive_ventilation_usd": 2450, "hospital_day_supplemental_oxygen_usd": 1828, "hospital_day_no_oxygen_usd": 1818 }, "not_reported": "The sources do not quantify this discovery project’s AI development, data licensing, compute, integration, maintenance, or human review costs." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Costs; Table III; Dosing and administration.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The model includes medicines, hospital care, later healthcare, and missed work. It excludes the costs of developing and operating the AI system." } }, "economics[].counterfactual": { "state": "value_present", "value": { "kind": "structured", "data": { "comparison": "Baricitinib plus standard care versus placebo plus the same standard care in the modeled COV-BARRIER population.", "base_case": "Uses numerical differences in mortality, recovery, and respiratory-support outcomes.", "mortality_only_scenario": "Limits treatment benefit to mortality reduction. Hospital costs rise rather than fall.", "AI_counterfactual": "No matched AI-versus-no-AI discovery scenario." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Model inputs; Analyses; Secondary Analyses.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The researchers compared baricitinib plus standard care with placebo plus the same care. Assuming only lower mortality reverses hospital savings to US$1,978 in additional costs per patient." } }, "economics[].currency_price_basis": { "state": "value_present", "value": { "kind": "structured", "data": { "currency": "USD", "jurisdiction": "United States", "price_basis": "The model uses U.S. dollar cost and payment inputs from 2020–2021. Some older care costs and future healthcare costs were inflated to 2020 USD; home-health and hospice reimbursement inputs use 2021 rates. The article does not establish that every input shares one uniform price year.", "conversion": "No currency conversion was used for these source figures." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Costs section, Table III, appendix input tables.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "U.S. dollars, with inputs from 2020–2021. Some costs were adjusted to 2020 prices, but the inputs do not share one price year." } }, "economics[].gross_net": { "state": "value_present", "value": { "kind": "structured", "data": { "hospital": "Net costs mean expenditures minus reimbursement. Base case: expenditure falls $2,436 and reimbursement falls $503, yielding a reported $1,932 net cost reduction.", "payer": "Incremental total costs include inpatient, postacute, future all-cause healthcare, and indirect costs. More survivors accumulate additional lifetime medical costs.", "exclusions": "No AI development or operating costs, long-COVID burden, COVID-related readmissions, or measured value from using freed beds for other patients." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Costs; Discussion limitations; appendix Tables S5–S6.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Hospital costs are net of reimbursement. Lifetime costs include healthcare and missed work. The model excludes long COVID, COVID-related readmissions, and the value of freed beds." } }, "economics[].measure_type": { "state": "value_present", "value": { "kind": "structured", "data": { "hospital": "Modeled change in hospital expenditure and net costs after reimbursement.", "payer": "Modeled incremental lifetime healthcare and indirect costs, and cost per quality-adjusted life-year.", "AI_boundary": "The model evaluates the treatment that researchers identified with AI assistance. It does not value AI separately." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Population and Perspective, Secondary Analyses, Table IV and supplementary Tables S5–S6.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Estimated hospital savings and the lifetime cost of additional health from treatment with baricitinib. The study assigns no separate monetary value to AI." } }, "economics[].period_unit": { "state": "value_present", "value": { "kind": "structured", "data": { "hospital": "Per modeled hospital admission, over the inpatient stay.", "payer": "Per modeled patient over a lifetime horizon.", "recurrence": "These are modeled per-patient differences, not recurring annual cash flows or a verified total across patients." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Population and Perspective; Model Framework.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Hospital costs cover one admission per patient. Costs to insurers cover each patient’s remaining lifetime." } }, "economics[].realization": { "state": "value_present", "value": { "kind": "structured", "data": { "state": "Modeled estimate", "basis": "Economic simulation using clinical trial effects and administrative cost inputs; no observed hospital ledger or cash saving was reported." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Methods and Secondary Analyses.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The amounts come from an economic model, without observed cash savings." } }, "economics[].recipient_payer": { "state": "value_present", "value": { "kind": "structured", "data": { "hospital_perspective": "Hospitals bear treatment costs and receive reimbursements. The estimated net saving reflects both components.", "payer_perspective": "Medicare, Medicaid, and private payers; the modeled mix is 52.9%, 19.2%, and 27.9%, respectively. The article also includes missed-work costs in its payer perspective.", "patients": "Clinical gains accrue to patients. The model excludes uninsured patients and does not establish individual out-of-pocket savings.", "AI_provider": "No payment, royalty, revenue, or profit to BenevolentAI is quantified by this analysis." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Population and Perspective; Table III; Costs.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Patients gain health. Hospitals pay for treatment and receive reimbursement from Medicare, Medicaid, and private insurers." } }, "economics[].return_payback": { "state": "value_present", "value": { "kind": "structured", "data": { "reported_status": "The source reports treatment costs and cost-effectiveness ratios, not a financial ROI, AI-project return, or exact payback date.", "aibd_calculated_health_value": { "name": "Incremental net monetary benefit", "perspective": "U.S. payer, per modeled patient over a lifetime", "formula": "Assumed USD value per quality-adjusted life-year × additional quality-adjusted life-years − additional lifetime costs", "source_inputs": { "base_additional_qalys": 0.6703, "base_additional_cost_usd": 17276, "mortality_only_additional_qalys": 0.6664, "mortality_only_additional_cost_usd": 17902 }, "scenarios": [ { "usd_per_qaly": 25000, "base_net_value_usd": -518.5, "mortality_only_net_value_usd": -1242, "assumption": "AIBD sensitivity scenario" }, { "usd_per_qaly": 50000, "base_net_value_usd": 16239, "mortality_only_net_value_usd": 15418, "assumption": "Threshold used in the original study" }, { "usd_per_qaly": 100000, "base_net_value_usd": 49754, "mortality_only_net_value_usd": 48738, "assumption": "AIBD sensitivity scenario" } ], "interpretation": "A valuation of additional health after additional costs, not a cash receipt or an estimate of AI’s separate contribution." }, "aibd_calculated_drug_cost_reference": { "formula": "$75.50 per 2-mg tablet × two tablets per day × 14 days", "full_course_reference_usd": 2114, "base_modeled_hospital_net_benefit_usd": 1932, "base_benefit_as_percent_of_reference": 91.4, "mortality_only_modeled_hospital_net_benefit_usd": -1978, "mortality_only_benefit_as_percent_of_reference": -93.6, "interpretation": "The model’s net hospital benefit divided by a full-course acquisition-cost benchmark. This is a reference ratio, not the model’s actual mean-cost ROI. Treatment could end earlier, dosing could differ, and acquisition prices could differ." }, "payback_assessment": { "accounting_horizon": "The modeled hospital admission", "base_case": "The model ends with lower net hospital costs.", "mortality_only_scenario": "The model ends with higher net hospital costs.", "cash_timing_status": "Purchase, payment, and savings dates are not reported, so an exact cash-payback day is not calculable from these totals." }, "estimation_confidence": "Arithmetic is reproducible from the published inputs. Economic conclusions depend on the model, valuation threshold, and clinical-effect assumptions." } }, "provenance": { "kind": "derivation", "basis": "Calculated the scenario values and reference ratios directly from the published economic model and appendix, with no invented AI attribution or cash-flow schedule.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "At US$50,000 per quality-adjusted life-year, AIBD values the additional health at US$16,239 per patient after costs: (50,000 × 0.6703) − 17,276. Cash returns and payment dates were not reported." } }, "economics[].valuation_method": { "state": "value_present", "value": { "kind": "structured", "data": { "method": "Linked inpatient, discharged, and recovered submodels based on COV-BARRIER effects, with duration of respiratory support rescaled using ACTT-2 data and U.S. administrative cost inputs. Hospital costs are days at each support level multiplied by daily cost, less reimbursement for net cost.", "discount_rate_per_year": 0.03, "sensitivity": "One-way and probabilistic sensitivity analyses (5,000 replications) for the payer model; a mortality-only scenario for both perspectives.", "payer_probabilistic_estimate": "Mean incremental cost $17,373 (95% interval −$3,300 to $38,306) and 0.674 QALYs (95% interval −0.096 to 1.441). These are model uncertainty results, not intervals from observed cash savings.", "source_dependency": "Some COV-BARRIER and Premier Healthcare Database inputs are data on file with Lilly and are not independently recoverable from the publication.", "funder": "Eli Lilly and Company.", "aibd_additional_calculations": "The return/payback field gives labeled net monetary benefit scenarios and a drug-cost reference ratio. These use the published model outputs; they are not new observations." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Model Framework, inputs, sensitivity analyses, Funding and appendix.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The researchers combined trial results with U.S. costs and applied a 3% annual discount rate. Hospital savings depend on clinical differences that were not all statistically conclusive." } }, "evaluations[].ai_uplift": { "state": "value_present", "value": { "kind": "structured", "data": { "reported_status": "No controlled estimate of additional value from AI is reported.", "aibd_inference": "Automated extraction and graph exploration probably reduced the manual work needed to assemble and examine relevant biomedical relationships. The published workflow supports this inference, but does not establish how much time was saved or whether the same drug would have been selected without AI.", "inference_confidence": "Moderate for assistance with information retrieval and prioritization. A numerical contribution to clinical benefit is not identifiable from these studies.", "documented_workflow_proxies": { "approximately_added_relationships": 40000, "proteins_in_initial_network": 556, "approved_AAK1_inhibitors_examined": 47, "high_affinity_AAK1_candidates": 6, "selected_drug": "Baricitinib", "iteration_timescale": "The developers describe iterations within hourly timeframes." }, "estimation_basis": "The original correspondence and methods paper report these workflow quantities. They measure the scale of the work, not an improvement against a human-only control.", "attribution_boundary": "No share of the clinical or economic result is assigned to AI." } }, "provenance": { "kind": "derivation", "basis": "Derived a qualitative contribution assessment from the documented mechanism and process counts. No causal percentage or time saving was invented.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0005", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The software probably reduced the work needed to assemble and examine relevant biomedical findings.", "editor_note": "Smith and colleagues report about 40,000 added relationships and a starting network of 556 proteins (reference 4). These quantities support an inference of substantial assistance, but the authors did not time a comparable search without AI." } }, "evaluations[].comparator": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "Baricitinib plus remdesivir versus placebo plus remdesivir, alongside supportive care.", "COV-BARRIER": "Baricitinib plus standard care versus placebo plus standard care; 79.3% received systemic corticosteroids at baseline and 18.9% remdesivir.", "RECOVERY": "Baricitinib plus usual care versus usual care alone. At baseline 95% received corticosteroids and 23% tocilizumab.", "economic_model": "Baricitinib plus standard care versus placebo plus standard care, using the COV-BARRIER population and treatment estimates.", "AI_comparison": "None of these trials randomized use of AI in the drug-discovery process." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Methods and baseline characteristics.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Patients received baricitinib or no baricitinib alongside the same background care. ACTT-2 gave remdesivir to both groups." } }, "evaluations[].design": { "state": "value_present", "value": { "kind": "structured", "data": { "evaluations": [ { "id": "ACTT-2", "design": "Randomized, double-blind, placebo-controlled trial." }, { "id": "COV-BARRIER", "design": "Randomized, double-blind, placebo-controlled phase 3 trial." }, { "id": "RECOVERY-baricitinib", "design": "Randomized, open-label, controlled platform trial." }, { "id": "Ohsfeldt-2021-economic-model", "design": "Simulation using clinical trial estimates, U.S. hospital cost data, published price and utility inputs, and sensitivity analyses." } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Methods.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Investigators randomly assigned treatment in all three trials. ACTT-2 and COV-BARRIER concealed assignments, while RECOVERY disclosed them. The economic study simulated costs and health." } }, "evaluations[].effect": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": { "recovery_rate_ratio": 1.16, "confidence_interval_95": [ 1.01, 1.32 ], "p": 0.03, "median_difference_days": -1, "mortality_hazard_ratio": 0.65, "mortality_confidence_interval_95": [ 0.39, 1.09 ], "mortality_conclusion": "Not statistically conclusive." }, "COV-BARRIER": { "primary_odds_ratio": 0.85, "primary_confidence_interval_95": [ 0.67, 1.08 ], "primary_p": 0.18, "primary_absolute_difference_percentage_points": -2.7, "mortality_hazard_ratio": 0.57, "mortality_confidence_interval_95": [ 0.41, 0.78 ], "mortality_nominal_p": 0.0018, "reported_mortality_absolute_difference_percentage_points": -5.0, "limitation": "Mortality did not meet multiplicity-controlled significance because the preceding primary endpoint failed." }, "RECOVERY": { "age_adjusted_mortality_rate_ratio": 0.87, "confidence_interval_95": [ 0.77, 0.99 ], "p": 0.028, "raw_rates": "514/4148 versus 546/4008; the paper rounds these to 12% and 14%.", "caution": "Do not call the difference in death counts (32) the number of lives saved, because arm sizes differ." }, "economic_model": { "hospital_base_net_cost_difference_usd": -1932, "hospital_mortality_only_cost_difference_usd": 1978, "lifetime_payer_base_cost_difference_usd": 17276 } } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at ACTT-2 Table 2; COV-BARRIER Table 2; RECOVERY Table 2; economic model Table IV, appendix Tables S5–S6.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Median recovery was 7 rather than 8 days in ACTT-2. Deaths by day 28 were 8% versus 13% in COV-BARRIER and 12% versus 14% in RECOVERY." } }, "evaluations[].exposure": { "state": "value_present", "value": { "kind": "structured", "data": { "basis": "Randomized participants are the matched denominators. Treatment-arm sizes are not counts of people proven to have benefited.", "ACTT-2": { "baricitinib_assigned": 515, "control": 518 }, "COV-BARRIER": { "baricitinib_assigned": 764, "control": 761, "safety_analysis_population": 1502 }, "RECOVERY": { "baricitinib_assigned": 4148, "usual_care": 4008 }, "economic_model": "Costs are per modeled eligible adult admission or patient lifetime, according to perspective; no observed number of hospital admissions generating the stated savings." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Participant flow, intention-to-treat populations, model population and perspective.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "ACTT-2 assigned 515 patients to baricitinib and 518 to control. COV-BARRIER assigned 764 and 761. RECOVERY assigned 4,148 and 4,008." } }, "evaluations[].id": { "state": "value_present", "value": { "kind": "structured", "data": { "evaluations": [ { "id": "ACTT-2", "registry": "NCT04401579" }, { "id": "COV-BARRIER", "registry": "NCT04421027" }, { "id": "RECOVERY-baricitinib", "registry": [ "NCT04381936", "ISRCTN50189673" ] }, { "id": "Ohsfeldt-2021-economic-model", "doi": "10.1016/j.clinthera.2021.09.016" } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Methods and article identifiers.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "ACTT-2, COV-BARRIER, RECOVERY, and Ohsfeldt and colleagues’ 2021 economic study." } }, "evaluations[].measurement_method": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "Clinical status on an eight-category ordinal scale; recovery was the first day in categories 1, 2, or 3 within 28 days. Trial staff collected clinical outcomes.", "COV-BARRIER": "Clinical progression on the NIAID ordinal scale, recorded mortality, and adjudicated safety events; study visits through day 28 and mortality assessment through day 60.", "RECOVERY": "Site staff completed follow-up forms at discharge, death, or day 28. Researchers also linked routinely collected healthcare and registry data.", "economic_model": "Three linked submodels covered inpatient care, discharge, and recovery. Model results are estimates rather than transactions or measured savings." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at ACTT-2 Outcomes; COV-BARRIER Procedures; RECOVERY Procedures; economics Model Framework.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Investigators used clinical assessments, follow-up forms, and health records. The economists combined trial findings with treatment prices and costs of care." } }, "evaluations[].metric": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": { "metric": "Time to recovery", "role": "Primary outcome", "period_days": 28, "baricitinib_median_days": 7, "control_median_days": 8 }, "COV-BARRIER": { "primary": "Progression to high-flow oxygen, noninvasive ventilation, invasive ventilation, ECMO, or death by day 28.", "primary_values_percent": [ 27.8, 30.5 ], "mortality_role": "Prespecified key secondary outcome", "mortality_days": 28, "baricitinib_deaths": 62, "baricitinib_randomized": 764, "control_deaths": 100, "control_randomized": 761 }, "RECOVERY": { "metric": "All-cause mortality", "role": "Primary outcome", "period_days": 28, "baricitinib_deaths": 514, "baricitinib_randomized": 4148, "control_deaths": 546, "control_randomized": 4008 }, "economic_model": { "metrics": [ "Incremental hospital expenditure", "Incremental reimbursement", "Incremental net hospital costs", "Incremental lifetime payer costs", "Quality-adjusted life-years", "Cost per quality-adjusted life-year" ] } } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Main outcomes and results tables; economic model Table IV and supplementary Table S6.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The primary outcomes were recovery time in ACTT-2, disease progression in COV-BARRIER, and deaths within 28 days in RECOVERY. The economists estimated costs and quality-adjusted life-years." } }, "evaluations[].reproduction_materials": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "NCT04401579; the publication links its protocol, supplementary appendix, disclosures, and data-sharing statement.", "COV-BARRIER": "NCT04421027; supplementary appendix 6 contains protocol details, the statistical plan, and additional outcomes.", "RECOVERY": "NCT04381936 and ISRCTN50189673; the article links the protocol/statistical plan and supplementary appendix; https://www.recoverytrial.net/ supplies trial documents.", "economic_model": "Published methods and an original DOCX appendix provide equations, inputs, and result Tables S1–S6. Some trial and hospital-cost inputs remain proprietary data on file. The full executable economic model and its underlying individual records are not supplied in the reviewed materials.", "AI_system": "The developers publish workflow methods, but not the full proprietary graph, a frozen January 2020 checkpoint, or an executable end-to-end reconstruction." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Supplementary material links, methods and data-sharing sections.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0004", "AIBD-0042-source-0010", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Trial registrations, protocols, analysis plans, and appendices are public. The complete historical AI system and economic model are unavailable, and some cost inputs are unpublished." } }, "evaluations[].setting_sample": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": { "randomized": 1033, "sites": 67, "countries": [ "United States", "Singapore", "South Korea", "Mexico", "Japan", "Spain", "United Kingdom", "Denmark" ], "enrollment_start": "2020-05-08", "enrollment_end": "2020-07-01", "population": "Hospitalized adults with COVID-19." }, "COV-BARRIER": { "randomized": 1525, "sites": 101, "countries_count": 12, "enrollment_start": "2020-06-11", "enrollment_end": "2021-01-15", "population": "Hospitalized adults with COVID-19 and an elevated inflammatory marker; patients already receiving invasive mechanical ventilation were excluded from the main trial." }, "RECOVERY": { "randomized": 8156, "country": "United Kingdom", "enrollment_start": "2021-02-02", "enrollment_end": "2021-12-29", "population": "Hospitalized patients with suspected or confirmed COVID-19; age at least two years under the protocol, with 33 children in the comparison." }, "economic_model": { "country": "United States", "population": "A modeled adult population based on COV-BARRIER; inpatient cost inputs from the Premier Healthcare Database, April 1–September 15, 2020. Uninsured patients excluded." } } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Study design, participants, Results, and economic model Population/Costs.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The trials enrolled hospitalized patients in 2020–2021: 1,033 in ACTT-2, 1,525 in COV-BARRIER, and 8,156 in RECOVERY. The economic study modeled insured U.S. adults." } }, "evaluations[].uncertainty_limitations": { "state": "value_present", "value": { "kind": "structured", "data": { "AI": "No controlled estimate of the AI workflow’s additional contribution; experts chose the candidate, and trials tested the medicine.", "ACTT-2": "Mortality was not the primary endpoint and its confidence interval included no effect. The trial occurred before corticosteroids became routine for these patients.", "COV-BARRIER": "The primary progression endpoint failed, leaving secondary mortality p values nominal under the prespecified hierarchy. Lilly sponsored, designed, and analyzed the trial.", "RECOVERY": "Open-label treatment, an age imbalance addressed in analysis, and a smaller mortality effect than earlier trials. Outcome data were available for more than 99% of randomized patients.", "economic_model": "Lilly funded the study. Base-case hospital savings depend on numerical clinical differences that were not all statistically conclusive; savings reverse in the mortality-only scenario. Some inputs are unpublished data on file. The model excludes long COVID, COVID-related readmissions, uninsured patients, and measured value from freed hospital capacity. Its U.S. costs may not generalize to individual hospitals or other countries." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Trial Discussion/Methods and economic model Discussion, Secondary Analyses, Funding/Conflicts.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "COV-BARRIER investigators did not establish less disease progression, which prevented confirmatory testing of mortality under their analysis plan. RECOVERY participants knew their treatment. Lilly funded COV-BARRIER and the economic study. None compared discovery with and without AI." } }, "external_references[]": { "state": "value_present", "value": { "kind": "structured", "data": { "trial_registrations": [ { "id": "NCT04401579", "trial": "ACTT-2" }, { "id": "NCT04421027", "trial": "COV-BARRIER" }, { "id": "NCT04381936", "trial": "RECOVERY" }, { "id": "ISRCTN50189673", "trial": "RECOVERY" } ], "scope_note": "No harm involving this system and episode is established by the reviewed sources. The prior Watson example involved another system." } }, "provenance": { "kind": "source_review", "basis": "Linked actual trial identifiers and removed an unrelated incident comparison.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "ACTT-2: NCT04401579. COV-BARRIER: NCT04421027. RECOVERY: NCT04381936 and ISRCTN50189673." } }, "follow_up[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "later": { "status": "persisted", "date": "2022-07-28", "note": "RECOVERY independently found lower 28-day mortality with baricitinib. WHO had recommended its use with corticosteroids for severe or critical COVID-19 in January 2022. The recommendation does not establish worldwide uptake.", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/" } } }, "provenance": { "kind": "source_review", "basis": "Recorded further clinical confirmation and kept policy guidance separate from demonstrated uptake.", "record_fields": [ "later" ], "source_refs": [ "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Oxford’s RECOVERY investigators independently confirmed lower mortality in July 2022." } }, "human_oversight": { "state": "value_present", "value": { "kind": "structured", "data": { "discovery": "Scientists iteratively queried the graph and assessed target binding, dose, and pharmacology before selecting the drug.", "clinical_testing": "Trial investigators, ethics committees, and monitoring arrangements governed treatment and follow-up.", "unreported": "The company’s internal software access controls and formal training requirements are not described." } }, "provenance": { "kind": "source_review", "basis": "The methods paper assigns enrichment, querying, and pharmacological assessment to named authors; trials describe their own oversight.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers chose the candidate drug. Clinicians prescribed treatment under trial protocols, with oversight from ethics committees and monitors." } }, "id": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "id": "AIBD-0042" } }, "provenance": { "kind": "source_review", "basis": "Preserved the existing case identifier.", "record_fields": [ "id" ], "source_refs": [], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "AIBD-0042" } }, "infrastructure_dependencies[]": { "state": "value_present", "value": { "kind": "structured", "data": { "data_sources": [ "KEGG", "Reactome", "STRING", "SIGNOR", "BioGRID", "ChEMBL", "CTD", "GWAS Catalog", "Gene Ontology", "OMIM", "Mondo" ], "software": "BenevolentAI knowledge graph and interactive network tools; SciPy is identified for enrichment analysis.", "unreported": [ "Compute hardware", "Cloud provider", "Hosting country", "Energy and water use", "Complete graph snapshot" ] } }, "provenance": { "kind": "source_review", "basis": "The methods paper names the data sources and software dependencies. Hardware and resource use are not documented.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The researchers used BenevolentAI’s graph, SciPy, and 11 biomedical resources, including KEGG, Reactome, and SIGNOR. Their methods paper does not identify the hardware or hosting provider." } }, "outcomes[].beneficiaries": { "state": "value_present", "value": { "kind": "structured", "data": { "personal": "People hospitalized with COVID-19, principally adults who required supplemental oxygen or ventilatory support. Trials evaluated groups rather than identifying which particular people survived because of treatment.", "organizational": "Hospitals that treated eligible patients; shorter recovery is a proxy for resource use, and the monetary benefit is modeled.", "societal": "Health systems with access to baricitinib and the capacity to give it with appropriate clinical supervision. Trials establish patient effects, not an independently measured population-wide effect." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Trial eligibility and results; economic model Population and Perspective.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Hospitalized patients with COVID-19 and the hospitals and health systems treating them." } }, "outcomes[].beneficiary_count": { "state": "value_present", "value": { "kind": "structured", "data": { "known_exposure": [ { "study": "ACTT-2", "randomized": 1033, "baricitinib_assigned": 515, "control_assigned": 518 }, { "study": "COV-BARRIER", "randomized": 1525, "baricitinib_assigned": 764, "control_assigned": 761 }, { "study": "RECOVERY", "randomized": 8156, "baricitinib_assigned": 4148, "control_assigned": 4008 } ], "actually_benefiting": "The individual number whose recovery or survival changed because of baricitinib is not directly observed. Randomized comparisons estimate an average treatment effect. Trial enrollment, treatment assignment, patients discharged, and people whose lives were saved are different counts.", "population_total": "No verified worldwide beneficiary count is established by these trial reports." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Results, participant flow and mortality tables.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The trials enrolled 10,714 participants in total, including controls. That total does not identify how many people benefited. The publications give no worldwide count." } }, "outcomes[].benefit_type": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "benefit_type": "Improved Physical Health", "taxonomy_note": "The clinical benefit is reduced mortality and faster recovery. AI assisted the researchers who proposed the drug for testing." } }, "provenance": { "kind": "source_review", "basis": "Assigned Improved Physical Health to the documented clinical benefit.", "record_fields": [ "benefit_type", "taxonomy_note" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Improved Physical Health" } }, "outcomes[].claim": { "state": "value_present", "value": { "kind": "structured", "data": { "benefits": [ { "id": "AIBD-0042-benefit-mortality", "title": "Fewer deaths", "statement": "In RECOVERY, 12% of patients receiving baricitinib died within 28 days, compared with 14% receiving usual care alone.", "benefit_type": "Improved Physical Health", "status": "reported", "beneficiaries": "Patients hospitalized with COVID-19.", "source_refs": [ "AIBD-0042-source-0002" ], "ratings": [ { "group": "personal", "depth": 5, "reason": "RECOVERY found lower 28-day mortality with baricitinib, a life-preserving benefit." } ], "evidence_word": "studied" }, { "id": "AIBD-0042-benefit-recovery", "title": "Faster recovery", "statement": "In ACTT-2, patients receiving remdesivir recovered in a median of 7 days with baricitinib, compared with 8 days with a placebo.", "benefit_type": "Improved Physical Health", "status": "reported", "beneficiaries": "Patients hospitalized with COVID-19.", "source_refs": [ "AIBD-0042-source-0005" ], "ratings": [ { "group": "personal", "depth": 2, "reason": "ACTT-2 found a shorter median recovery time in a randomized comparison, a measurable improvement in patients’ health." } ], "evidence_word": "studied" } ] } }, "provenance": { "kind": "source_review", "basis": "Separated the two measured clinical benefits already documented in the record. RECOVERY supports lower mortality; ACTT-2 supports faster recovery. The trials evaluated treatment effects, not the incremental value of AI-assisted drug selection. Each benefit has its own source references and beneficiary assessment. The existing hospital and societal assessments remain in the record.", "record_fields": [ "account", "benefit_type" ], "source_refs": [ "AIBD-0042-source-0002", "AIBD-0042-source-0005" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Fewer deaths: In RECOVERY, 12% of patients receiving baricitinib died within 28 days, compared with 14% receiving usual care alone. Faster recovery: In ACTT-2, patients receiving remdesivir recovered in a median of 7 days with baricitinib, compared with 8 days with a placebo." } }, "outcomes[].distribution": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "The high-flow oxygen or noninvasive ventilation subgroup had median recovery of 10 versus 18 days (recovery rate ratio 1.51, 95% CI 1.10–2.08), compared with 7 versus 8 days overall. Subgroup estimates have lower precision.", "COV-BARRIER": "Effects were examined by age, sex, region, baseline respiratory support, symptom duration, remdesivir, and corticosteroids. The large mortality difference among patients receiving high-flow oxygen or noninvasive ventilation is a secondary subgroup result subject to the trial’s multiplicity limit. The trial did not establish a positive benefit in every participating country.", "RECOVERY": "No evidence of treatment-effect differences across the six prespecified subgroups (age, sex, ethnicity, symptom duration, respiratory support, corticosteroid use; interaction p values >0.18). This is not proof that all subgroups benefit equally.", "economic_access": "The U.S. cost model excluded uninsured patients because out-of-pocket cost data were unavailable. It does not establish savings in other jurisdictions." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at ACTT-2 Table 2; COV-BARRIER Figures 2–3; RECOVERY Figure 3; economic model Costs.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "ACTT-2 patients on high-flow oxygen or noninvasive ventilation recovered in a median of 10 rather than 18 days. This subgroup estimate is less precise. RECOVERY found no clear differences among subgroups." } }, "outcomes[].duration": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "Clinical outcomes followed through day 28.", "COV-BARRIER": "Primary and key secondary outcomes through day 28; exploratory mortality through day 60, with 79/764 deaths versus 116/761 (hazard ratio 0.62, 95% CI 0.47–0.83).", "RECOVERY": "The reported main outcomes were assessed at 28 days; a planned six-month analysis is not evidence of a six-month result.", "economic_model": "Hospital estimates cover the admission; payer estimates extend over a modeled lifetime. The lifetime benefit is projected, not observed follow-up." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Trial Outcomes/Results; economic model Population and Perspective.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The trials followed patients for 28 days. COV-BARRIER investigators also examined deaths through day 60 in an exploratory analysis. The economists projected costs and health over a lifetime." } }, "outcomes[].id": { "state": "value_present", "value": { "kind": "structured", "data": { "benefit_ids": [ "AIBD-0042-benefit-mortality", "AIBD-0042-benefit-recovery" ], "stakeholder_assessment_ids": { "personal": "AIBD-0042-outcome-personal", "organizational": "AIBD-0042-outcome-organizational", "societal": "AIBD-0042-outcome-societal" } } }, "provenance": { "kind": "curatorial", "basis": "Assigned separate stable identifiers to mortality and recovery while retaining the earlier stakeholder-assessment identifiers.", "record_fields": [], "source_refs": [], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "AIBD-0042-benefit-mortality, AIBD-0042-benefit-recovery" } }, "outcomes[].stakeholder_group": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "personal", "depth": 5, "reason": "Fewer patients died with baricitinib in COV-BARRIER and RECOVERY. RECOVERY confirmed lower 28-day mortality as its primary outcome in 8,156 patients." }, { "group": "organizational", "depth": 2, "reason": "ACTT-2 reported median recovery of 7 days with baricitinib and 8 days without it. That could free hospital capacity. A separate economic model estimated lower hospital costs, but neither establishes realized staffing or cash savings." }, { "group": "societal", "depth": 4, "reason": "WHO recommended baricitinib with corticosteroids for severe or critical COVID-19. That recommendation supports wider access to a treatment that reduced deaths in trials. Worldwide uptake and population-wide effects were not measured." } ] } }, "provenance": { "kind": "source_review", "basis": "Preserved the personal, organizational, and societal targets with explicit distinctions in their support.", "record_fields": [ "ratings" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "People, organizations, and society." } }, "outcomes[].status": { "state": "value_present", "value": { "kind": "structured", "data": { "personal": "Observed treatment benefit in trials", "organizational": "Potential capacity benefit and modeled financial result, not measured realized savings", "societal": "International treatment recommendation observed; population-wide effects not measured" } }, "provenance": { "kind": "source_review", "basis": "Separated observed patient outcomes from projected or unmeasured wider effects.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0007", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Patient benefits were observed. Hospital savings were modeled. WHO recommended wider treatment. These studies did not measure population-wide effects." } }, "outcomes[].tradeoffs": { "state": "value_present", "value": { "kind": "structured", "data": { "ACTT-2": "Serious adverse events occurred in 16.0% with baricitinib plus remdesivir and 21.0% with remdesivir alone. Venous thromboembolism occurred in 4.1% versus 3.1%, a difference of 1.0 percentage point (95% CI −1.3 to 3.3).", "COV-BARRIER": "Among 750 baricitinib recipients and 752 placebo recipients in the safety population, serious adverse events occurred in 110 (15%) versus 135 (18%), serious infections in 64 (9%) versus 74 (10%), and venous thromboembolism in 20 (3%) versus 19 (3%). These findings concern short treatment and follow-up and do not establish absence of rare or long-term harms.", "RECOVERY": "No significant excess in the prespecified infection, thrombotic, or bleeding outcomes was detected. Investigators reported 12 suspected serious adverse reactions attributed to baricitinib, including four serious infections other than COVID-19, three bowel perforations, and two pulmonary emboli.", "costs": "The model estimated lower hospital net costs only under its base-case assumptions. Its mortality-only scenario estimated higher hospital costs, and the lifetime payer model estimated higher overall cost because more patients survived." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Trial safety results; RECOVERY Results and appendix p. 61 reference; economic model Secondary Analyses.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Serious adverse events affected 15% of baricitinib recipients and 18% of controls in COV-BARRIER. RECOVERY investigators reported 12 suspected serious reactions. Short follow-up leaves rare and long-term harms uncertain." } }, "outcomes[].unexpected": { "state": "value_present", "value": { "kind": "structured", "data": { "assessment": "Expected in purpose: researchers deliberately sought an existing medicine that could help people with COVID-19. Mortality reduction was a prespecified trial outcome, although it was secondary in COV-BARRIER. A failed progression endpoint and favorable mortality result are different findings, not grounds for calling the benefit accidental." } }, "provenance": { "kind": "source_review", "basis": "Reviewed the original source at Methods paper Introduction and workflow; trial prespecified Outcomes.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0004", "AIBD-0042-source-0001", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers deliberately sought a treatment, and investigators planned to measure mortality." } }, "places[]": { "state": "value_present", "value": { "kind": "structured", "data": { "location": "Multicountry trials (ACTT-2, COV-BARRIER, and RECOVERY): Argentina; Brazil; Denmark; Germany; India; Italy; Japan; Mexico; Russia; Singapore; South Korea; Spain; United Kingdom; United States", "developer_country": "United Kingdom", "deployer_country": "United Kingdom", "developer_jurisdictions": [ "United Kingdom" ], "case_jurisdictions": [ "Argentina", "Brazil", "Denmark", "Germany", "India", "Italy", "Japan", "Mexico", "Russia", "Singapore", "South Korea", "Spain", "United Kingdom", "United States" ], "location_label": "Trial locations", "location_note": "Trial countries, with results pooled across sites. RECOVERY was conducted in the United Kingdom. Puerto Rico is included under the United States.", "ai_operation": "BenevolentAI research team, with author affiliations in London and Brooklyn. The exact compute location is not reported.", "clinical_evaluators": [ "ACTT-2: NIAID, United States", "COV-BARRIER: Eli Lilly, United States", "RECOVERY: University of Oxford, United Kingdom" ] } }, "provenance": { "kind": "source_review", "basis": "Separated the AI operator’s jurisdiction from trial locations and evaluators. Headquarters does not establish compute location.", "record_fields": [ "location", "developer_country", "deployer_country" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "BenevolentAI operated in the United Kingdom. Trials ran in Argentina, Brazil, Denmark, Germany, India, Italy, Japan, Mexico, Russia, Singapore, South Korea, Spain, the United Kingdom, and the United States." } }, "purpose": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "purpose": "Researchers searched for existing drugs that might treat COVID-19, using a biomedical knowledge graph that combines databases with information extracted from scientific papers." } }, "provenance": { "kind": "source_review", "basis": "The original investigators state their intended search for a COVID-19 treatment.", "record_fields": [ "purpose" ], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers sought existing medicines that could treat COVID-19." } }, "record_dates": { "state": "value_present", "value": { "kind": "structured", "data": { "discovery": { "start": "2020-01", "end": "2020-01", "precision": "month" }, "first_ai_publication": "2020-02-04", "actt2_enrollment": { "start": "2020-05-08", "end": "2020-07-01" }, "cov_barrier_enrollment": { "start": "2020-06-11", "end": "2021-01-15" }, "recovery_enrollment": { "start": "2021-02-02", "end": "2021-12-29" }, "case_date": "2021-09-01", "case_date_basis": "Original publication of the COV-BARRIER mortality findings", "date_precision": "day", "first_database_revision": "2026-10-20", "database_discovery_note": "No separate discovery date is recorded.", "last_reviewed": "2026-10-25" } }, "provenance": { "kind": "source_review", "basis": "Separated discovery, study periods, publication, and editorial dates.", "record_fields": [ "date_observed", "date_basis", "date_precision", "revisions" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers identified baricitinib in January 2020 and reported it on 4 February 2020. COV-BARRIER mortality results appeared on 1 September 2021." } }, "replication.attempts[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "BenevolentAI researchers combined a biomedical knowledge graph with expert review to select baricitinib.", "Clinical investigators separately tested the drug alongside the care available in 2020 and 2021." ], "must_hold": [ "The clinical comparisons concern hospitalized patients and the eligibility criteria of each trial.", "Background treatment differed: remdesivir in ACTT-2, and mostly corticosteroids in COV-BARRIER and RECOVERY.", "Reproducing the AI selection would also require the January 2020 graph, software configuration, and researchers’ decisions. The complete snapshot is not public." ], "echoes": [ { "date": "2021-09-01", "setting": "COV-BARRIER: 1,525 adults. Lower mortality was a secondary finding after an unsuccessful primary comparison.", "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/" }, { "date": "2022-07-28", "setting": "RECOVERY: 8,156 patients in 159 U.K. hospitals. Lower 28-day mortality was the primary finding. These were further tests of the drug, not repetitions of the AI discovery process.", "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/" } ] } } }, "provenance": { "kind": "source_review", "basis": "Replaced the WHO recommendation with a genuine subsequent clinical trial. Different trial protocols and endpoints remain explicit.", "record_fields": [ "replication" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "COV-BARRIER and RECOVERY both found fewer deaths with baricitinib. The trials tested the medicine, without repeating the AI discovery process." } }, "replication.conditions": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "BenevolentAI researchers combined a biomedical knowledge graph with expert review to select baricitinib.", "Clinical investigators separately tested the drug alongside the care available in 2020 and 2021." ], "must_hold": [ "The clinical comparisons concern hospitalized patients and the eligibility criteria of each trial.", "Background treatment differed: remdesivir in ACTT-2, and mostly corticosteroids in COV-BARRIER and RECOVERY.", "Reproducing the AI selection would also require the January 2020 graph, software configuration, and researchers’ decisions. The complete snapshot is not public." ], "echoes": [ { "date": "2021-09-01", "setting": "COV-BARRIER: 1,525 adults. Lower mortality was a secondary finding after an unsuccessful primary comparison.", "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/" }, { "date": "2022-07-28", "setting": "RECOVERY: 8,156 patients in 159 U.K. hospitals. Lower 28-day mortality was the primary finding. These were further tests of the drug, not repetitions of the AI discovery process.", "outcome": "recurred", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/" } ] } } }, "provenance": { "kind": "source_review", "basis": "Distinguished further drug trials from replication of the discovery workflow.", "record_fields": [ "replication" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Testing the treatment again requires comparable patients and background care. Repeating the discovery requires the original graph, software configuration, and researchers’ decisions." } }, "revision_history[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "editor": "AIBD" }, { "date": "2026-10-20", "change": "Paired with AI Incident Database incident 225 (medium confidence); severity not given by the source.", "source": "https://incidentdatabase.ai/cite/225", "editor": "AIBD" }, { "date": "2026-10-20", "change": "Recategorised from specific harm prevented to 1 because the sources show an AI-generated repurposing hypothesis confirmed at cohort level by randomized trials rather than a specific harm prevented for an identifiable beneficiary, and, on the fact-checker's other findings, made the ACTT-2 quote verbatim, relabelled the Lancet correspondence letter as “other”, used the exact source titles, replaced the unsupported “hospitals worldwide” with the trial sponsors as deployers, and stated in the summary and account that COV-BARRIER missed its primary endpoint and that its 28-day mortality reduction was a secondary endpoint.", "source": "Richardson P, et al. Lancet 2020 (https://europepmc.org/article/PMC/PMC7137985); Kalil AC, et al. N Engl J Med 2021 (https://europepmc.org/article/PMC/PMC7745180); Marconi VC, et al. Lancet Respir Med 2021 (https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/)", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Added what the benefit is measured against, what happened since from the cited sources.", "source": "https://europepmc.org/article/PMC/PMC7745180; https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Improved Physical Health under Trustible's full taxonomy of 23 types; it was Improved Knowledge Discovery among the six first used.", "source": "https://trustible.ai/post/a-framework-for-measuring-the-benefits-of-ai/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Country of the developer and the deployer added from the organization's own site or a registry.", "source": "https://www.benevolent.com/contact-us/; https://en.wikipedia.org/wiki/National_Institute_of_Allergy_and_Infectious_Diseases", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the replication conditions rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "editor": "AIBD" }, { "date": "2026-10-23", "change": "Personal benefit depth changed from 4 to 5 for the reported life-preserving outcome.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Traced the AI hypothesis, recovery and mortality findings, and regulatory actions to their original sources. Put COV-BARRIER first for the mortality claim, preserved the failed primary endpoint and added the multiplicity limitation. Removed the generic FDA drug-list page from active evidence because it adds no case-specific support beyond the retained authorization. Distinguished NIAID-led ACTT-2 from the Lilly-sponsored COV-BARRIER analysis.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Rewrote the short title and summary in plain language, preserving the reported benefit and material limits. Added the named publication or institution to the reviewed lead source for homepage display.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Identified the AI system or feature in the supporting summary, or stated that the source does not name the software. Preserved the reported benefit and material qualifications.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Traced the system identity beyond the original case report. Added product names and supporting original sources, preserving historical-version limits and any inference in the source notes.", "source": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Shortened the homepage supporting statement to one sentence naming the system and its contribution; retained detailed outcomes and qualifications in the case account and source notes.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Revised the title and subtitle for natural phrasing, with named researchers, shorter dependencies, and less compressed terminology; preserved the system identities and evidence qualifications.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Added the developers’ original methods paper, identified the SVO component and January 2020 workflow, and clarified the scientists’ role in selecting baricitinib. The exact model release remains unreported.", "source": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "editor": "AIBD" }, { "date": "2026-10-24", "change": "Revised the homepage subtitle for concise database prose, with courtesy titles for named beneficiaries and numbered clauses for the Coates drug-selection account. Preserved the reported evidence and material uncertainty.", "editor": "AIBD" }, { "date": "2026-10-25", "change": "Corrected the developer jurisdiction to its contemporaneous U.K. headquarters and replaced the London-only location with the 14 countries in ACTT-2 and COV-BARRIER. Changed the chronology from the FDA authorization date to September 1, 2021, the original publication date of the COV-BARRIER mortality findings. Trial countries do not imply a separately demonstrated benefit in every country.", "source": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "editor": "AIBD" }, { "date": "2026-10-25", "change": "Rewrote the account and resolved all mandatory fields. Added RECOVERY and the original economic model, corrected the AI operator, distinguished projected wider benefits from observed patient outcomes, replaced WHO guidance counted as replication, and removed an unrelated Watson incident. Missing model and attribution details remain explicit. Reordered the sources by their contribution and remapped their identifiers.", "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "editor": "AIBD, Codex-assisted review" }, { "date": "2026-10-25", "change": "Added historical control from original corporate filings, documented funding and scientific-resource relationships, a labeled strategic assessment, and derived economic scenarios. Preserved the unreported exact model version, with a documented configuration proxy. Distinguished reported findings, calculations, and inferences throughout the assessment.", "source": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "editor": "AIBD, Codex-assisted review" } ] } }, "provenance": { "kind": "source_review", "basis": "Recorded the corrections with prior versions recoverable in repository history.", "record_fields": [ "revisions" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Ownership, funding, and economic estimates were added on 25 October 2026. Earlier versions remain available." } }, "sector": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "domain": "Biomedicine / infectious disease therapeutics", "segment": "Hospitalized COVID-19 patients requiring supplemental oxygen or ventilatory support" } }, "provenance": { "kind": "source_review", "basis": "The application concerns drug discovery and treatment of hospitalized patients.", "record_fields": [ "domain", "segment" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Biomedicine and hospital treatment of infectious disease." } }, "sources[].access_version": { "state": "value_present", "value": { "kind": "structured", "data": { "reviewed_on": "2026-10-25", "versions_and_access": [ { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC7302766/", "finding": "June 18, 2020 correction restores two omitted BenevolentAI authors." }, { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8425787/", "finding": "September 8, 2021 correction changes the baricitinib number at risk on day 14 in Figure 2A to 684." }, { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "supplements": "Original and final protocols, analysis plans, supplementary methods/results, and Figure S2 were inspected. Patient-level trial data were not reanalyzed." }, { "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC4780864/", "status": "Full original text, Figure 3 and its description, and methods inspected through Europe PMC XML. Supplementary Table S1 remained inaccessible." }, { "source_url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "status": "Original source inspected.", "locator": "Operational highlights, baricitinib bullet.", "reviewed_on": "2026-10-24" }, { "source_url": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "source_ref": "AIBD-0042-source-0004", "publication_date": "2021-07-28", "accessed_on": "2026-10-24", "status": "Original publisher text inspected.", "locator": "Materials and Methods; Results, opening paragraphs and figure 3; Discussion." }, { "source_url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "source_ref": "AIBD-0042-source-0009", "accessed_on": "2026-10-25", "language": "English", "status": "Original developer webpage inspected for corporate geography.", "locator": "About BenevolentAI, final paragraph: London headquarters.", "publication_date": "2020-09-14" }, { "source_ref": "AIBD-0042-source-0001", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "access_url": "https://rcastoragev2.blob.core.windows.net/bf64edd09bbf67d7978734149d4c838a/PMC8409066.pdf", "accessed_on": "2026-10-25", "status": "Original article text and table 1 inspected for country coverage and publication date.", "locator": "Table 1, p. 1411; publication date, p. 1407." }, { "source_ref": "AIBD-0042-source-0005", "source_url": "https://europepmc.org/article/PMC/PMC7745180", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7745180/fullTextXML", "accessed_on": "2026-10-25", "status": "Full original article XML inspected after intermittent PMC browser checks.", "locator": "Methods, Design, paragraph 2 (section sc1.1)." }, { "source_ref": "AIBD-0042-source-0001", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8409066/fullTextXML", "accessed_on": "2026-10-25", "status": "Original article text inspected through Europe PMC XML." }, { "source_ref": "AIBD-0042-source-0005", "source_url": "https://europepmc.org/article/PMC/PMC7745180", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7745180/fullTextXML", "accessed_on": "2026-10-25", "status": "Original article text inspected through Europe PMC XML." }, { "source_ref": "AIBD-0042-source-0003", "source_url": "https://europepmc.org/article/PMC/PMC7137985", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC7137985/fullTextXML", "accessed_on": "2026-10-25", "status": "Original article text inspected through Europe PMC XML." }, { "source_ref": "AIBD-0042-source-0004", "source_url": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC8356560/fullTextXML", "accessed_on": "2026-10-25", "status": "Original article text inspected through Europe PMC XML." }, { "source_ref": "AIBD-0042-source-0002", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "access_url": "https://www.ebi.ac.uk/europepmc/webservices/rest/PMC9333998/fullTextXML", "accessed_on": "2026-10-25", "status": "Original article text inspected through Europe PMC XML." }, { "source_ref": "AIBD-0042-source-0010", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8487786/", "accessed_on": "2026-10-25", "status": "Original article and supplemental table S6 inspected.", "supplement": "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx" }, { "source_ref": "AIBD-0042-source-0002", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "publication_date": "2022-07-28", "issue_date": "2022-07-30", "verification_url": "https://eprints.gla.ac.uk/292687/", "finding": "The university repository identifies July 28, 2022 as the original online publication date. July 30 is the issue date.", "accessed_on": "2026-10-25" }, { "source_ref": "AIBD-0042-source-0002", "source_url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9521822/", "accessed_on": "2026-10-25", "finding": "The October 2022 erratum states that the RECOVERY appendix was corrected on September 29, 2022. It does not identify the amendment, so no assertion about its effect on the findings is made." }, { "source_url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "accessed_on": "2026-10-25", "status": "Original scanned filing inspected, including a visual check of note 22.", "locator": "p. 43, note 22; p. 40, note 18." }, { "source_url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "accessed_on": "2026-10-25", "status": "Original scanned statement and relevant shareholder pages inspected.", "locator": "Share capital, p. 2; shareholder details, pp. 4–6 (unnumbered leaves)." }, { "source_url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "accessed_on": "2026-10-25", "status": "Original scanned return and share-rights pages inspected.", "locator": "pp. 2–3." }, { "source_url": "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "accessed_on": "2026-10-25", "status": "Original counsel announcement inspected.", "locator": "¶¶ 1–3." }, { "source_url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "accessed_on": "2026-10-25", "status": "Original developer announcement inspected.", "locator": "Opening paragraph; funding-purpose paragraph; About BenevolentAI." }, { "source_url": "https://www.genome.jp/kegg/kegg1.html", "accessed_on": "2026-10-25", "status": "Original resource overview inspected.", "locator": "KEGG overview and project history." }, { "source_url": "https://doi.org/10.1093/nar/gkz949", "accessed_on": "2026-10-25", "status": "Original resource paper inspected.", "locator": "Author affiliations; database access and download description." }, { "source_url": "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf", "accessed_on": "2026-10-25", "status": "Original report downloaded and read after browser access failed.", "locator": "p. 38, “Relating to Our Shareholder.”" } ], "limitations": [ "No trial randomized the discovery process to AI versus no AI. The full proprietary graph and ranking run are unavailable.", "COV-BARRIER was sponsor-designed and sponsor-analyzed. Its mortality inference is nominal under the prespecified testing hierarchy.", "The cited recommendation establishes guidance, not measured worldwide adoption or a count of lives saved.", "The original Sorrell et al. article was read. Its linked supplementary Table S1 was identified, but the download returned 403 or a verification page and the supplementary archive timed out; no unavailable table contents were inferred." ] } }, "provenance": { "kind": "source_review", "basis": "Preserved corrections and access limitations and recorded the new primary-source inspections.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The discovery letter, COV-BARRIER figure, and RECOVERY appendix have corrections. Sorrell’s supplementary table and the January 2022 WHO guideline PDF were inaccessible, although the article and WHO announcement were available." } }, "sources[].citation_locator": { "state": "value_present", "value": { "kind": "structured", "data": { "main_works": [ { "kind": "peer-reviewed study", "citation": "Marconi VC, Ramanan AV, de Bono S, et al. (2021). Efficacy and safety of baricitinib for the treatment of hospitalised adults with COVID-19 (COV-BARRIER). Lancet Respir Med. Results and Figure 2; supplementary appendix 6, pp. 9–14, including the prespecified testing hierarchy.", "title": "Efficacy and safety of baricitinib for the treatment of hospitalised adults with COVID-19 (COV-BARRIER): a randomised, double-blind, parallel-group, placebo-controlled phase 3 trial", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "date": "2021-09-01", "written_by": "evaluator", "quote": "The 28-day all-cause mortality was 8% (n=62) for baricitinib and 13% (n=100) for placebo (hazard ratio [HR] 0·57 [95% CI 0·41–0·78]; nominal p=0·0018)" }, { "kind": "peer-reviewed study", "citation": "RECOVERY Collaborative Group. “Baricitinib in Patients Admitted to Hospital with COVID-19 (RECOVERY): A Randomised, Controlled, Open-Label, Platform Trial and Updated Meta-Analysis.” The Lancet 400 (2022): 359–368. Table 2, Figure 3, and Data sharing. https://doi.org/10.1016/S0140-6736(22)01109-6.", "title": "Baricitinib in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial and updated meta-analysis", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "date": "2022-07-28", "written_by": "independent" }, { "kind": "other", "citation": "Richardson P, Griffin I, Tucker C, Smith D, Oechsle O, Phelan A, Rawling M, Savory E, Stebbing J (2020). Baricitinib as potential treatment for 2019-nCoV acute respiratory disease. Lancet.", "title": "Baricitinib as potential treatment for 2019-nCoV acute respiratory disease", "url": "https://europepmc.org/article/PMC/PMC7137985", "date": "2020-02-04", "written_by": "makers", "quote": "BenevolentAI's knowledge graph is a large repository of structured medical information, including numerous connections extracted from scientific literature by machine learning." }, { "kind": "peer-reviewed study", "citation": "Daniel P. Smith et al. “Expert-Augmented Computational Drug Repurposing Identified Baricitinib as a Treatment for COVID-19.” Frontiers in Pharmacology 12 (July 28, 2021): 709856. https://doi.org/10.3389/fphar.2021.709856.", "title": "Expert-Augmented Computational Drug Repurposing Identified Baricitinib as a Treatment for COVID-19", "url": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "date": "2021-07-28", "written_by": "makers" }, { "kind": "peer-reviewed study", "citation": "Kalil AC, Patterson TF, Mehta AK, et al. (2021). Baricitinib plus Remdesivir for Hospitalized Adults with Covid-19. N Engl J Med. Table 2 and trial oversight; supplementary appendix, pp. 19–23; protocol NCT04401579.", "title": "Baricitinib plus Remdesivir for Hospitalized Adults with Covid-19", "url": "https://europepmc.org/article/PMC/PMC7745180", "date": "2020-12-11", "written_by": "independent", "quote": "median time to recovery of 7 days (95% confidence interval [CI], 6 to 8), as compared with 8 days (95% CI, 7 to 9)" }, { "kind": "press release", "citation": "U.S. Food and Drug Administration. “Coronavirus (COVID-19) Update: FDA Authorizes Drug Combination for Treatment of COVID-19.” November 19, 2020. Authorization, evidentiary standard, and ACTT-2 results.", "title": "FDA Authorizes Drug Combination for Treatment of COVID-19", "url": "https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-authorizes-drug-combination-treatment-covid-19", "date": "2020-11-19", "written_by": "independent", "quote": "the drug baricitinib, in combination with remdesivir, for the treatment of suspected or laboratory confirmed COVID-19" }, { "kind": "report", "citation": "World Health Organization (2022). WHO recommends two new drugs to treat COVID-19.", "title": "WHO recommends two new drugs to treat COVID-19", "url": "https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "date": "2022-01-14", "written_by": "independent", "quote": "The first drug, baricitinib, is strongly recommended for patients with severe or critical COVID-19." }, { "kind": "report", "citation": "BenevolentAI. “BenevolentAI Unaudited Preliminary Results for the Year Ended 31 December 2022,” March 16, 2023, Operational highlights, baricitinib bullet.", "title": "BenevolentAI Unaudited Preliminary Results for the Year Ended 31 December 2022", "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "date": "2023-03-16", "written_by": "makers" }, { "kind": "press release", "citation": "BenevolentAI. “BenevolentAI’s platform-derived hypothesis for COVID-19 treatment validated in US NIAID randomised control trial.” September 14, 2020. “About BenevolentAI,” headquarters statement.", "title": "BenevolentAI’s platform-derived hypothesis for COVID-19 treatment validated in US NIAID randomised control trial", "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "date": "2020-09-14", "written_by": "makers" }, { "kind": "peer-reviewed study", "citation": "Robert Ohsfeldt et al. “Cost-Effectiveness of Baricitinib Compared With Standard of Care: A Modeling Study in Hospitalized Patients With COVID-19 in the United States.” Clinical Therapeutics 43, no. 11 (2021): 1877–1893.e4. Tables IV–V and supplemental table S6. https://doi.org/10.1016/j.clinthera.2021.09.016.", "title": "Cost-Effectiveness of Baricitinib Compared With Standard of Care: A Modeling Study in Hospitalized Patients With COVID-19 in the United States", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8487786/", "date": "2021-10-04", "written_by": "evaluator" } ], "metadata_sources": [ { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Annual Report and Consolidated Financial Statements, 31 December 2019. Filed December 30, 2020.", "locator": "p. 43, note 22; p. 40, note 18.", "date": "2019-12-31", "contribution": "Historical company control and share-class rights immediately before the January 2020 research.", "relationship": "Original statutory company filing. The control statement is from the company’s accounts, not an inference from later ownership.", "access_note": "Original scanned filing inspected, including a visual check of note 22." }, { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Confirmation Statement CS01, October 15, 2019.", "locator": "Share capital, p. 2; shareholder details, pp. 4–6 (unnumbered leaves).", "date": "2019-10-15", "contribution": "The historical voting-share counts used for the labeled AIBD calculation.", "relationship": "Original company filing, with the ordinary and preferred share rights stated explicitly.", "access_note": "Original scanned statement and relevant shareholder pages inspected." }, { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Return of Allotment SH01, December 20, 2019. Filed January 8, 2020.", "locator": "pp. 2–3.", "date": "2019-12-20", "contribution": "A capital check showing that the new G2 growth shares did not carry ordinary voting rights before an exit.", "relationship": "Original company filing. This checks share-class rights, not changes in every beneficial owner.", "access_note": "Original scanned return and share-rights pages inspected." }, { "url": "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "citation": "Covington & Burling LLP. “Covington Advises BenevolentAI on $90M Investment from Temasek.” October 15, 2019.", "locator": "¶¶ 1–3.", "date": "2019-10-15", "contribution": "A participant’s confirmation of Singaporean investment in the developer.", "relationship": "The law firm advised BenevolentAI on the transaction. It is a firsthand participant, not an independent financial evaluator.", "access_note": "Original counsel announcement inspected." }, { "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "citation": "BenevolentAI. “BenevolentAI Announces $90 Million Investment from Temasek.” September 17, 2019.", "locator": "Opening paragraph; funding-purpose paragraph; About BenevolentAI.", "date": "2019-09-17", "contribution": "The original funding announcement, its date, and the intended company-wide uses of the capital.", "relationship": "The recipient’s own announcement. It does not allocate funding to the later baricitinib work.", "access_note": "Original developer announcement inspected." }, { "url": "https://www.genome.jp/kegg/kegg1.html", "citation": "Kanehisa Laboratories. “KEGG Database.”", "locator": "KEGG overview and project history.", "contribution": "The Japanese origin of a biomedical resource that the discovery researchers used.", "relationship": "Resource developers’ account of its history. Geographic reuse is inferred by combining it with the baricitinib methods paper.", "access_note": "Original resource overview inspected." }, { "url": "https://doi.org/10.1093/nar/gkz949", "citation": "Luana Licata et al. “SIGNOR 2.0, the SIGnaling Network Open Resource 2.0: 2019 Update.” Nucleic Acids Research 48, no. D1 (2020): D504–D510. https://doi.org/10.1093/nar/gkz949.", "locator": "Author affiliations; database access and download description.", "date": "2019-10-29", "contribution": "An Italy-led resource, with U.K. collaborators, that the discovery researchers reused.", "relationship": "Original resource paper. It establishes its producers and access conditions, not server or patient-data routes.", "access_note": "Original resource paper inspected." }, { "url": "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf", "citation": "Temasek Holdings. Temasek Overview 2019.", "locator": "p. 38, “Relating to Our Shareholder.”", "date": "2019", "contribution": "The investor’s government ownership and its stated separation from portfolio-company business decisions.", "relationship": "The investor’s own institutional account. It does not establish government control of BenevolentAI or the research.", "access_note": "Original report downloaded and read after browser access failed." } ] } }, "provenance": { "kind": "source_review", "basis": "Preserved the full clinical and discovery citations and added precise locators for corporate and geographic evidence.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "References 1–10 cover discovery, trials, and costs. References 11–18 cover ownership, funding, and geography." } }, "sources[].id": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "role": "Ten main works in reviewed order, followed by identifiers for the metadata sources" }, "provenance": { "kind": "source_review", "basis": "Kept the main evidence identifiers stable and used direct URLs to identify additional metadata sources.", "record_fields": [ "evidence" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "AIBD-0042-source-0001 through AIBD-0042-source-0010, plus eight supplementary references." } }, "sources[].language": { "state": "value_present", "value": { "kind": "structured", "data": { "language": "English", "translation_used": false, "scope": "All retained source passages were read in English. Source titles and quotations preserve original spelling." } }, "provenance": { "kind": "source_review", "basis": "Inspected the retained English-language publications.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "English. No translation was needed." } }, "sources[].relationship": { "state": "value_present", "value": { "kind": "structured", "data": { "sources": [ { "source_ref": "AIBD-0042-source-0001", "relationship": "Original trial by clinicians and the drug sponsor; not an independent evaluation of the AI discovery process." }, { "source_ref": "AIBD-0042-source-0002", "relationship": "Oxford-led trial, independent of BenevolentAI and the drug manufacturer. Baricitinib came from routine NHS stock." }, { "source_ref": "AIBD-0042-source-0003", "relationship": "BenevolentAI authors and an adviser report their own hypothesis." }, { "source_ref": "AIBD-0042-source-0004", "relationship": "Original methods paper from the developers. It adds technical detail, not an independent evaluation of AI’s clinical benefit." }, { "source_ref": "AIBD-0042-source-0005", "relationship": "NIAID controlled trial design and implementation. Lilly and Gilead supplied medicines and provided protocol input, but did not fund ACTT-2." }, { "source_ref": "AIBD-0042-source-0006", "relationship": "Original regulator statement; its efficacy figures derive from ACTT-2." }, { "source_ref": "AIBD-0042-source-0007", "relationship": "Original WHO recommendation based on trial synthesis, not another clinical experiment." }, { "source_ref": "AIBD-0042-source-0008", "relationship": "Developer’s retrospective identification, not independent verification of clinical benefit." }, { "source_ref": "AIBD-0042-source-0009", "relationship": "Original developer statement. Its clinical claims are not counted as additional verification of the trials." }, { "source_ref": "AIBD-0042-source-0010", "relationship": "Economic study funded by Eli Lilly, with company-employed coauthors. Trial inputs and cost assumptions are modeled, not another clinical experiment." } ], "caution": "An original source is not necessarily independent. FDA and WHO assess treatment evidence; they do not independently measure the contribution of the AI discovery workflow.", "metadata_sources": [ { "source_ref": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "relationship": "Original statutory company filing. The control statement is from the company’s accounts, not an inference from later ownership." }, { "source_ref": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "relationship": "Original company filing, with the ordinary and preferred share rights stated explicitly." }, { "source_ref": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "relationship": "Original company filing. This checks share-class rights, not changes in every beneficial owner." }, { "source_ref": "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "relationship": "The law firm advised BenevolentAI on the transaction. It is a firsthand participant, not an independent financial evaluator." }, { "source_ref": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "relationship": "The recipient’s own announcement. It does not allocate funding to the later baricitinib work." }, { "source_ref": "https://www.genome.jp/kegg/kegg1.html", "relationship": "Resource developers’ account of its history. Geographic reuse is inferred by combining it with the baricitinib methods paper." }, { "source_ref": "https://doi.org/10.1093/nar/gkz949", "relationship": "Original resource paper. It establishes its producers and access conditions, not server or patient-data routes." }, { "source_ref": "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf", "relationship": "The investor’s own institutional account. It does not establish government control of BenevolentAI or the research." } ] } }, "provenance": { "kind": "source_review", "basis": "Reviewed authors, funding, and document purpose separately from study design.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The developers supplied four sources. Four were independent, and two involved the drug sponsor. Lilly funded COV-BARRIER and the economic study. Manufacturers supplied ACTT-2 medicines and protocol advice, without funding the trial." } }, "sources[].upstream_sources": { "state": "value_present", "value": { "kind": "structured", "data": { "reviewed_on": "2026-10-25", "lead_source_ref": "AIBD-0042-source-0001", "retained_source_order": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010" ], "origins": [ { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "contribution": "Lilly-funded COV-BARRIER: mortality and disease progression in 1,525 adults. Mortality was a secondary finding after an unsuccessful primary comparison.", "locator": "Results, Figure 2; supplementary appendix 6, pp. 9–14 and Figure S2. Table 1, p. 1411; publication date, p. 1407.", "relationship": "Original trial by clinicians and the drug sponsor; not an independent evaluation of the AI discovery process.", "source_refs": [ "AIBD-0042-source-0001" ], "display_name": "The Lancet Respiratory Medicine" }, { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "source_refs": [ "AIBD-0042-source-0002" ], "display_name": "The Lancet", "contribution": "RECOVERY: an independent randomized trial of 8,156 patients, with 28-day mortality as its primary outcome.", "locator": "Table 2; Figure 3; Methods; Declaration of interests; Data sharing.", "relationship": "Oxford-led trial, independent of BenevolentAI and the drug manufacturer. Baricitinib came from routine NHS stock." }, { "url": "https://europepmc.org/article/PMC/PMC7137985", "contribution": "The original February 2020 proposal to test baricitinib, using BenevolentAI’s graph.", "locator": "Lancet 395:e30–e31, text, figure, and references, corrected June 18, 2020.", "relationship": "BenevolentAI authors and an adviser report their own hypothesis.", "source_refs": [ "AIBD-0042-source-0003" ], "display_name": "The Lancet" }, { "url": "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full", "source_refs": [ "AIBD-0042-source-0004" ], "locator": "Materials and Methods; Results, opening paragraphs and figure 3; Discussion.", "display_name": "Frontiers in Pharmacology", "contribution": "The developers explain the graph, algorithms, and scientists’ role in selecting baricitinib.", "relationship": "Original methods paper from the developers. It adds technical detail, not an independent evaluation of AI’s clinical benefit." }, { "url": "https://europepmc.org/article/PMC/PMC7745180", "contribution": "NIAID-led ACTT-2: recovery time in 1,033 adults. Drug suppliers provided medicines and protocol input but did not fund the trial.", "locator": "Table 2; Trial Oversight; supplementary appendix, pp. 19–23; NCT04401579. Methods, Design, paragraph 2.", "relationship": "NIAID controlled trial design and implementation. Lilly and Gilead supplied medicines and provided protocol input, but did not fund ACTT-2.", "source_refs": [ "AIBD-0042-source-0005" ] }, { "url": "https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-authorizes-drug-combination-treatment-covid-19", "contribution": "FDA’s November 2020 authorization and its evidentiary basis. The decision concerns the drug, not the AI discovery method.", "locator": "November 19, 2020 announcement, authorization link and ACTT-2 paragraphs.", "relationship": "Original regulator statement; its efficacy figures derive from ACTT-2.", "source_refs": [ "AIBD-0042-source-0006" ] }, { "url": "https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "contribution": "WHO’s January 2022 recommendation. Guidance is not another clinical trial or a measure of worldwide uptake.", "locator": "January 14, 2022 statement, baricitinib recommendation and guideline link.", "relationship": "Original WHO recommendation based on trial synthesis, not another clinical experiment.", "source_refs": [ "AIBD-0042-source-0007" ] }, { "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/", "contribution": "The developer identifies the January 2020 discovery system as the Benevolent Platform.", "locator": "Operational highlights, baricitinib bullet.", "relationship": "Developer’s retrospective identification, not independent verification of clinical benefit.", "source_refs": [ "AIBD-0042-source-0008" ], "display_name": "BenevolentAI" }, { "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "contribution": "BenevolentAI’s London headquarters during the 2020 work.", "locator": "About BenevolentAI, final paragraph: London headquarters.", "relationship": "Original developer statement. Its clinical claims are not counted as additional verification of the trials.", "source_refs": [ "AIBD-0042-source-0009" ], "display_name": "BenevolentAI" }, { "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8487786/", "source_refs": [ "AIBD-0042-source-0010" ], "display_name": "Clinical Therapeutics", "contribution": "A Lilly-funded model of U.S. treatment costs. Its hospital saving depends on the scenario and does not measure savings attributable to AI.", "locator": "Tables IV–V; supplemental table S6; Participants and Methods; Conflicts of Interest.", "relationship": "Economic study funded by Eli Lilly, with company-employed coauthors. Trial inputs and cost assumptions are modeled, not another clinical experiment." } ], "claim_chains": [ { "claim": "Mortality and failure of the primary endpoint", "path": [ "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "Supplementary appendix 6: Statistical analysis, pp. 9–10; Select secondary outcomes, pp. 11–12; Figure S2, p. 14" ], "finding": "The failure of the primary endpoint left no alpha for confirmatory testing of key secondary endpoints. The mortality p-value is nominal." }, { "claim": "Recovery time", "path": [ "https://europepmc.org/article/PMC/PMC7745180", "Table 2; supplementary appendix and original/final protocols" ], "finding": "ACTT-2 compared baricitinib plus remdesivir with placebo plus remdesivir. Its mortality comparison was not statistically conclusive." }, { "claim": "AI contribution", "path": [ "https://europepmc.org/article/PMC/PMC7137985", "Figure and original reference 6: Sorrell et al., Structure 24 (2016):401–411, doi:10.1016/j.str.2015.12.015", "https://pmc.ncbi.nlm.nih.gov/articles/PMC4780864/" ], "finding": "The correspondence is the origin of the AI-assisted prioritization claim. Sorrell et al. (2016), Figure 3B and the ligand-interaction results, supply original laboratory evidence of baricitinib binding to AAK1. That biochemical result does not establish the proposed antiviral mechanism or the clinical contribution of the AI discovery process." }, { "claim": "Regulatory follow-up", "path": [ "https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-authorizes-drug-combination-treatment-covid-19", "https://www.fda.gov/media/143822/download", "https://www.who.int/news/item/14-01-2022-who-recommends-two-new-drugs-to-treat-covid-19", "https://iris.who.int/bitstream/handle/10665/351006/WHO-2019-nCoV-therapeutics-2022.1-eng.pdf" ], "finding": "The current FDA letter is a later reissuance with authorization history. The original WHO January 2022 PDF was identified but returned 403; the dated WHO announcement was read." }, { "claim": "System identity", "path": [ "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-unaudited-preliminary-results-year-ended-31-december-2022/" ], "finding": "Benevolent Platform, using its machine-learning-built knowledge graph and AAK1 inhibitor screening" }, { "claim": "Underlying workflow and named component", "path": [ "https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2021.709856/full" ], "finding": "The authors name SVO as a component and describe a workflow that combined computational analysis with expert selection. They do not establish an exact model release." }, { "claim": "Developer jurisdiction and geographic scope", "path": [ "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentais-platform-derived-hypothesis-covid-19-treatment-validated-us-niaid-randomised-control-trial/", "https://pmc.ncbi.nlm.nih.gov/articles/PMC8409066/", "https://europepmc.org/article/PMC/PMC7745180", "https://europepmc.org/article/PMC/PMC7137985" ], "finding": "The original correspondence and company statement identify London, United Kingdom, as the discovery team’s base. COV-BARRIER and ACTT-2 together cover 14 countries, with Puerto Rico included under the United States. These are trial settings, not separately demonstrated benefits in every country. The AI operator is BenevolentAI’s research team; clinical trial investigators evaluated the drug. Specific cross-border relationships are assessed separately." }, { "claim": "Further clinical confirmation", "path": [ "https://pmc.ncbi.nlm.nih.gov/articles/PMC9333998/", "Table 2 and Figure 3" ], "finding": "RECOVERY tested baricitinib against usual care. It did not replicate the AI discovery process." }, { "claim": "Economic estimates", "path": [ "https://pmc.ncbi.nlm.nih.gov/articles/PMC8487786/", "Tables IV–V", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx", "Supplemental table S6" ], "finding": "The model projects payer costs and hospital net costs. It does not measure cash savings or isolate an AI contribution." }, { "claim": "Historical ownership and control", "path": [ "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "p. 43, note 22", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0" ], "finding": "The accounts identify Kenneth Mulvany as the controller at year-end 2019. A separate AIBD calculation describes his voting-share fraction in the October register." }, { "claim": "Cross-border company funding", "path": [ "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek" ], "finding": "The company and its counsel report investment from Singapore-headquartered Temasek. The amount is company-wide and is not a measured benefit or research budget." }, { "claim": "Cross-border scientific resources", "path": [ "AIBD-0042-source-0004", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949" ], "finding": "Named resource use is documented. The country-to-country description is an AIBD inference from the resource producers and the discovery team." }, { "claim": "AIBD monetary scenarios", "path": [ "AIBD-0042-source-0010", "Tables IV–V", "https://ars.els-cdn.com/content/image/1-s2.0-S0149291821003891-mmc1.docx", "Supplemental table S6" ], "finding": "The return/payback assessment derives net monetary benefit and a full-course cost reference ratio from published inputs. Neither is an observed cash return or an AI-specific return." } ], "limitations": [ "No trial randomized the discovery process to AI versus no AI. The full proprietary graph and ranking run are unavailable.", "COV-BARRIER was sponsor-designed and sponsor-analyzed. Its mortality inference is nominal under the prespecified testing hierarchy.", "The cited recommendation establishes guidance, not measured worldwide adoption or a count of lives saved.", "The original Sorrell et al. article was read. Its linked supplementary Table S1 was identified, but the download returned 403 or a verification page and the supplementary archive timed out; no unavailable table contents were inferred.", "The authors’ measured recovery times do not establish realized hospital staffing or cash savings.", "The FDA source is a news release about an authorization, not a finding about AI effectiveness." ], "excluded_or_consolidated_sources": [ { "url": "https://www.fda.gov/drugs/emergency-preparedness-drugs/coronavirus-covid-19-drugs", "reason": "Generic, changing drug directory; no additional claim in the current case requires it." }, { "url": "https://incidentdatabase.ai/cite/225", "reason": "IBM Watson for Oncology is a different system and a different clinical use. Its incident does not document a harm in this baricitinib discovery or treatment episode." } ], "review_scope": "Reviewed all required fields for this discovery episode. The ten main works establish discovery, clinical effects, institutional decisions, and modeled treatment costs. Additional corporate filings, transaction accounts, and resource histories support ownership and geographic assessment. AIBD calculations and interpretations are labeled separately from reported findings.", "homepage_source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0003" ], "metadata_sources": [ { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Annual Report and Consolidated Financial Statements, 31 December 2019. Filed December 30, 2020.", "locator": "p. 43, note 22; p. 40, note 18.", "date": "2019-12-31", "contribution": "Historical company control and share-class rights immediately before the January 2020 research.", "relationship": "Original statutory company filing. The control statement is from the company’s accounts, not an inference from later ownership.", "access_note": "Original scanned filing inspected, including a visual check of note 22." }, { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Confirmation Statement CS01, October 15, 2019.", "locator": "Share capital, p. 2; shareholder details, pp. 4–6 (unnumbered leaves).", "date": "2019-10-15", "contribution": "The historical voting-share counts used for the labeled AIBD calculation.", "relationship": "Original company filing, with the ordinary and preferred share rights stated explicitly.", "access_note": "Original scanned statement and relevant shareholder pages inspected." }, { "url": "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "citation": "BenevolentAI Limited. Return of Allotment SH01, December 20, 2019. Filed January 8, 2020.", "locator": "pp. 2–3.", "date": "2019-12-20", "contribution": "A capital check showing that the new G2 growth shares did not carry ordinary voting rights before an exit.", "relationship": "Original company filing. This checks share-class rights, not changes in every beneficial owner.", "access_note": "Original scanned return and share-rights pages inspected." }, { "url": "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "citation": "Covington & Burling LLP. “Covington Advises BenevolentAI on $90M Investment from Temasek.” October 15, 2019.", "locator": "¶¶ 1–3.", "date": "2019-10-15", "contribution": "A participant’s confirmation of Singaporean investment in the developer.", "relationship": "The law firm advised BenevolentAI on the transaction. It is a firsthand participant, not an independent financial evaluator.", "access_note": "Original counsel announcement inspected." }, { "url": "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "citation": "BenevolentAI. “BenevolentAI Announces $90 Million Investment from Temasek.” September 17, 2019.", "locator": "Opening paragraph; funding-purpose paragraph; About BenevolentAI.", "date": "2019-09-17", "contribution": "The original funding announcement, its date, and the intended company-wide uses of the capital.", "relationship": "The recipient’s own announcement. It does not allocate funding to the later baricitinib work.", "access_note": "Original developer announcement inspected." }, { "url": "https://www.genome.jp/kegg/kegg1.html", "citation": "Kanehisa Laboratories. “KEGG Database.”", "locator": "KEGG overview and project history.", "contribution": "The Japanese origin of a biomedical resource that the discovery researchers used.", "relationship": "Resource developers’ account of its history. Geographic reuse is inferred by combining it with the baricitinib methods paper.", "access_note": "Original resource overview inspected." }, { "url": "https://doi.org/10.1093/nar/gkz949", "citation": "Luana Licata et al. “SIGNOR 2.0, the SIGnaling Network Open Resource 2.0: 2019 Update.” Nucleic Acids Research 48, no. D1 (2020): D504–D510. https://doi.org/10.1093/nar/gkz949.", "locator": "Author affiliations; database access and download description.", "date": "2019-10-29", "contribution": "An Italy-led resource, with U.K. collaborators, that the discovery researchers reused.", "relationship": "Original resource paper. It establishes its producers and access conditions, not server or patient-data routes.", "access_note": "Original resource paper inspected." }, { "url": "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf", "citation": "Temasek Holdings. Temasek Overview 2019.", "locator": "p. 38, “Relating to Our Shareholder.”", "date": "2019", "contribution": "The investor’s government ownership and its stated separation from portfolio-company business decisions.", "relationship": "The investor’s own institutional account. It does not establish government control of BenevolentAI or the research.", "access_note": "Original report downloaded and read after browser access failed." } ], "interpretation_boundary": "Reported findings, direct calculations, and AIBD inferences are distinguished in the affected fields. A completed assessment does not imply that the exact proprietary model or a causal AI contribution is known." } }, "provenance": { "kind": "source_review", "basis": "Traced clinical, discovery, policy, economic, corporate, and geographic statements to their separate origins. Recorded the basis of derived calculations and interpretations.", "record_fields": [ "evidence" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0002", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0009", "AIBD-0042-source-0010", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI0NjgzNDI0M2FkaXF6a2N4/document?format=pdf&download=0", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI1NDE0MDc3M2FkaXF6a2N4/document?format=pdf&download=0", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949", "https://tr19.temasekreview.com.sg/downloads/Temasek-Review-2019-Overview.pdf" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The FDA and WHO based their conclusions on clinical trials. BenevolentAI’s announcements repeat its discovery account and the trial results." } }, "strategic_assessments[]": { "state": "value_present", "value": { "kind": "structured", "data": { "assessments": [ { "id": "AIBD-0042-strategic-uk-preparedness", "assessment_type": "AIBD qualitative inference", "assessed_on": "2026-10-25", "perspective": "United Kingdom public health preparedness and biomedical research capacity", "domain": "Pandemic preparedness", "reported_status": "The sources document the research network and treatment results. They do not quantify a change in national capability or an advantage attributable to AI.", "documented_basis": "A U.K. discovery team used international biomedical resources after Singaporean investment in the company. Researchers subsequently tested the drug in U.S.-led, multinational, and U.K. trials.", "aibd_inference": "This sequence illustrates a route from existing medicines to clinical tests during a new disease outbreak. Its value for U.K. preparedness depends on access to international knowledge, expert pharmacology, trial capacity, and medicines.", "capability_change": "The researchers produced a testable treatment hypothesis and the trials established a clinical benefit. The record does not identify how much sooner the hypothesis emerged because of AI.", "operational_dependency": "The published trial protocols administer a medicine without using BenevolentAI’s discovery software. AIBD therefore infers that continuing access to that software was not necessary for the treatment regimen that the trials evaluated.", "confidence": "Moderate for this qualitative assessment. No numerical national advantage, time saving, or strategic score is assigned.", "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949" ] } ] } }, "provenance": { "kind": "derivation", "basis": "Prepared a named, dated strategic assessment from the documented institutions and protocols. Marked the capability interpretation as AIBD inference.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0005", "AIBD-0042-source-0001", "AIBD-0042-source-0002", "https://find-and-update.company-information.service.gov.uk/company/09781806/filing-history/MzI4NjgwMDAzNWFkaXF6a2N4/document?format=pdf&download=0", "https://www.benevolent.com/news-and-media/press-releases-and-in-media/benevolentai-announces-90-million-investment-temasek/", "https://www.cov.com/en/news-and-insights/news/2019/10/covington-advises-benevolentai-on-90m-investment-from-temasek", "https://www.genome.jp/kegg/kegg1.html", "https://doi.org/10.1093/nar/gkz949" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "This work may strengthen the United Kingdom’s ability to find treatments during a new outbreak.", "editor_note": "The inference rests on the U.K. team’s January 2020 proposal and the subsequent international trials (references 1–5). Those studies establish a useful treatment, but give no estimate of how much earlier AI made it available." } }, "summary": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "summary": "Researchers used the Benevolent Platform to identify baricitinib as a candidate treatment for COVID-19.", "account": "BenevolentAI researchers reported in February 2020 that their knowledge graph had helped them identify baricitinib, an arthritis drug, as a possible treatment for COVID-19. They proposed clinical testing after reviewing its effects on inflammation and proteins involved in viral entry. In ACTT-2, 1,033 hospitalized adults received remdesivir with either baricitinib or a placebo. Median recovery took 7 days with baricitinib and 8 days with the placebo.\n\nCOV-BARRIER enrolled 1,525 adults. By day 28, 8% of patients receiving baricitinib had died, compared with 13% receiving a placebo. Both groups received standard care. Mortality was a secondary outcome. The trial did not meet its primary outcome of reduced disease progression, so its mortality result was not conclusive under the planned procedure for multiple comparisons.\n\nThe subsequent RECOVERY trial enrolled 8,156 patients. Mortality, its primary outcome, was 12% with baricitinib and 14% with usual care alone." } }, "provenance": { "kind": "source_review", "basis": "Separated clinical results from the unmeasured additional contribution of AI.", "record_fields": [ "summary", "account" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers used the Benevolent Platform to identify baricitinib as a candidate treatment for COVID-19." } }, "systems[].autonomy_permissions": { "state": "value_present", "value": { "kind": "structured", "data": { "operating_mode": "Interactive analysis directed by scientists", "selection_authority": "Researchers chose the candidate drug and proposed clinical testing.", "clinical_authority": "No autonomous prescribing or treatment execution is described.", "technical_permissions": "API permissions, access controls, and escalation rules are not reported." } }, "provenance": { "kind": "source_review", "basis": "The documented workflow preserves candidate selection and clinical decisions with people.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Scientists directed the analysis and selected the drug. Clinicians controlled treatment." } }, "systems[].component_role": { "state": "value_present", "value": { "kind": "structured", "data": { "roles": [ "Extract biomedical relationships from literature", "Retrieve relationships from a knowledge graph", "Display and analyze protein networks" ], "human_decision": "Scientists selected baricitinib after pharmacological assessment." } }, "provenance": { "kind": "source_review", "basis": "Methods and Author Contributions distinguish automated analysis from scientist selection.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The system extracted biomedical relationships, searched the graph, and analyzed networks of proteins." } }, "systems[].deployment_configuration": { "state": "value_present", "value": { "kind": "structured", "data": { "period": "January 2020", "workflow": "Coronavirus literature was added to the graph, followed by iterative queries and network analysis.", "unreported": [ "Hardware", "Hosting location", "Model checkpoint", "Exact application release", "Frozen graph snapshot", "Compute budget" ], "reported_status": "The exact model release and frozen graph snapshot are not reported.", "aibd_configuration_proxy": "BenevolentAI’s January 2020 coronavirus-augmented knowledge graph, combined with SVO relationship extraction, graph-pattern queries, and interactive protein-network analysis.", "proxy_basis": "Date and components are documented in the original methods paper. They identify a historical workflow, not an executable version or checkpoint.", "proxy_confidence": "High for the documented period and components; insufficient for exact computational reproduction.", "documented_components": { "relationship_extraction": "SVO extracts relationships from sentences using an unsupervised, rule-based approach.", "network_analysis": "Graph-pattern queries and an interactive protein-network tool with Louvain community detection.", "initial_network": { "proteins": 556, "interactions": 2286, "modules": 16 }, "human_review": "Researchers considered target affinity, free plasma concentration, pharmacokinetics, and dose when selecting baricitinib." } } }, "provenance": { "kind": "source_review", "basis": "Recorded a historically documented configuration proxy while leaving the exact model-version field unreported.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Researchers added relationships from coronavirus papers to the knowledge graph. They used the Louvain algorithm to group connected proteins and examined the networks with interactive tools." } }, "systems[].inputs_outputs": { "state": "value_present", "value": { "kind": "structured", "data": { "inputs": [ "Biomedical databases", "Scientific literature about coronaviruses, proteins, pathways, and drugs" ], "outputs": [ "Relationships among biological entities", "Protein interaction networks", "Candidate drugs for expert review" ], "language": "English-language search examples are documented; the full language coverage of the corpus is not reported.", "patient_data": "The discovery workflow describes biomedical literature and databases, not individual trial-patient records." } }, "provenance": { "kind": "source_review", "basis": "Read Materials and Methods and Results in the original methods paper.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The system drew from biomedical literature and databases to produce relationships and protein networks that researchers used to assess candidate drugs." } }, "systems[].model_name": { "state": "value_present", "value": { "kind": "structured", "data": { "named_component": "SVO", "role": "Rule-based extraction of sentence-level relationships", "other_methods": [ "Graph pattern queries", "Louvain community detection", "Protein interaction network analysis" ], "single_model": "No single model represents the entire workflow." } }, "provenance": { "kind": "source_review", "basis": "The methods paper names the component and algorithms.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "SVO, a rule-based model that extracts relationships from sentences." } }, "systems[].model_version": { "state": "not_reported", "provenance": { "kind": "source_review", "basis": "The original methods paper does not identify an exact model release, checkpoint, or frozen January 2020 graph snapshot.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "reason": "The original methods paper does not identify an exact model release, checkpoint, or frozen January 2020 graph snapshot.", "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The configuration used in January 2020.", "editor_note": "Smith and colleagues describe the January 2020 work but give no release identifier for SVO (reference 4, p. 4)." } }, "systems[].product_name": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "system": "Benevolent Platform" } }, "provenance": { "kind": "source_review", "basis": "The original workflow is identified retrospectively as the Benevolent Platform. This is a product name, not a model release.", "record_fields": [ "system" ], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0008", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Benevolent Platform" } }, "systems[].provider": { "state": "value_present", "value": { "kind": "structured", "data": { "application_developer": "BenevolentAI", "application_operator": "BenevolentAI research team", "named_component_provider": "BenevolentAI" } }, "provenance": { "kind": "source_review", "basis": "The authors describe their own application and research workflow.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0003", "AIBD-0042-source-0004" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "BenevolentAI" } }, "taxonomy_crosswalks[]": { "state": "value_present", "value": { "kind": "structured", "data": { "mapping": [ { "taxonomy": "Trustible AI Benefits Framework", "version": "May 1, 2024; AIBD local taxonomy snapshot", "category": "Improved Physical Health", "relationship": "Exact category assignment", "rationale": "Lower mortality and faster recovery concern patients’ physical health." } ], "other_taxonomies": "No crosswalk to an incident or risk taxonomy is asserted." } }, "provenance": { "kind": "curatorial", "basis": "Applied the existing local benefit taxonomy, independently of the depth scale.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "The benefit matches Trustible’s “Improved Physical Health” category, published on 1 May 2024." } }, "title": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "title": "Baricitinib, identified with BenevolentAI, reduced COVID-19 deaths in trials", "headline": "An arthritis drug helped patients with severe COVID-19" } }, "provenance": { "kind": "source_review", "basis": "The title describes the drug benefit and identifies the discovery relationship.", "record_fields": [ "title", "headline" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0003", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "An arthritis drug helped patients with severe COVID-19" } }, "version": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "a8e1b0052ad9c7eff39b20f5389ce8ba29abb827c2271690fcac5264ab7db26f" ], "role": "SHA-256 of normalized case data" }, "provenance": { "kind": "curatorial", "basis": "Recomputed after integrating the sourced metadata and labeled estimates.", "record_fields": [ "revisions" ], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0006", "AIBD-0042-source-0007", "AIBD-0042-source-0008", "AIBD-0042-source-0004", "AIBD-0042-source-0009", "AIBD-0042-source-0002", "AIBD-0042-source-0010" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Revision a8e1b005, reviewed on 25 October 2026." } }, "workflow": { "state": "value_present", "value": { "kind": "text", "text": "BenevolentAI scientists added coronavirus literature to their graph, queried it, examined protein networks, and assessed candidate drugs. Clinical investigators subsequently tested baricitinib in separate trials." }, "provenance": { "kind": "source_review", "basis": "Read the discovery workflow and trial oversight sections.", "record_fields": [], "source_refs": [ "AIBD-0042-source-0001", "AIBD-0042-source-0005", "AIBD-0042-source-0003", "AIBD-0042-source-0004", "AIBD-0042-source-0002" ], "reviewed_on": "2026-10-25" }, "review_state": "reviewed", "coverage": "complete", "display": { "answer": "Scientists searched the graph, examined protein networks, and assessed medicines. Clinical investigators then tested baricitinib." } } }, "summary": { "total": 91, "addressed": 91, "known": 90, "reviewed": 91, "record_mapped": 0, "not_assessed": 0, "complete": true, "by_state": { "value_present": 90, "not_reported": 1, "unresolved": 0, "not_assessed": 0, "not_applicable": 0, "withheld": 0, "source_unavailable": 0 } } } }