{ "id": "AIBD-0184", "title": "DWP's Universal Credit advances machine-learning model saved an estimated £4.4 million from 2022–23 to 2024–25", "summary": "The National Audit Office reports that DWP's model, in use since May 2022, flags high-risk advance requests for caseworker review, with estimated savings of £4.4 million and about three times the h...", "date_observed": "2025-10-22", "date_basis": "unassessed", "date_precision": "unassessed", "benefit_status": null, "source_collection": null, "source_entry": null, "account": "The National Audit Office (NAO) published “Tackling benefit overpayments due to fraud and error” on 22 October 2025. It states that since May 2022 the Department for Work & Pensions (DWP) has used a machine learning model. The model flags potentially fraudulent claims for Universal Credit advances. DWP estimates it saved £4.4 million from 2022–23 to 2024–25. The model scores new claims and refers those above a threshold to a caseworker. Decisions are made by DWP staff. DWP cannot disaggregate the model's costs from wider spending on data analytics. The NAO reports DWP's finding that the model is around three times more effective at identifying fraud risk than a randomized control group sample. It reports a median payment delay of one day for referred requests. DWP's July 2025 fairness analysis for 2024–25 showed older claimants (age groups 45 to 54 and above) and non-UK nationals over-referred for review. Fairness analysis was possible on only one of nine protected characteristics. Four further models were in development. Civil Service World (Jim Dunton, 22 October 2025) reported the NAO's warnings on data infrastructure. It quoted Transformation Minister Andrew Western saying that the one machine learning model has made the department significantly more effective.", "category": "general benefit", "unexpected": false, "purpose": "Assess the fraud risk of Universal Credit advance requests and refer high-risk requests to a caseworker for review.", "system": "Universal Credit advances machine learning model", "model_version": null, "developer": "Department for Work & Pensions (DWP)", "deployer": "Department for Work & Pensions (DWP)", "location": null, "domain": "Benefit fraud and error", "segment": "Universal Credit advances", "benefit_type": "Reduced Risk", "taxonomy_note": null, "registry_ids": {}, "ratings": [ { "group": "organizational", "depth": null, "reason": "The saving is an aggregate departmental estimate over three years, not an attributable outcome." }, { "group": "societal", "depth": null, "reason": "No outcome for claimants or taxpayers beyond the aggregate estimate is documented. Referred claimants bear an added evidence burden." } ], "ai_role": { "status": "assessed", "statement": "The NAO describes a model trained on historical fraud and error data that scores advance requests and refers those above a threshold. Every payment decision is made by a caseworker, so the model's contribution is prioritization. Developer and deployer are the same public body (DWP). The NAO, an independent auditor, wrote the report." }, "evidence": [ { "kind": "report", "citation": "National Audit Office, 'Tackling benefit overpayments due to fraud and error', 22 October 2025, paragraphs 15-16 and 3.4-3.15", "title": "Tackling benefit overpayments due to fraud and error", "url": "https://www.nao.org.uk/reports/tackling-benefit-overpayments-due-to-fraud-and-error/", "date": "2025-10-22", "written_by": "independent", "quote": "Since May 2022, DWP has used a machine learning model to flag potentially fraudulent claims for UC advances, saving an estimated £4.4 million." }, { "kind": "news report", "citation": "Jim Dunton, Civil Service World, 22 October 2025", "title": "Digital and data challenges could hamper DWP's fraud strategy, NAO warns", "url": "https://www.civilserviceworld.com/news/article/digital-and-data-challenges-could-hamper-dwps-fraud-strategy-nao-warns", "date": "2025-10-22", "written_by": "independent", "quote": "our one machine learning model has made us significantly more effective" }, { "kind": "report", "citation": "Department for Work & Pensions (2026). Effectiveness Assessment of Universal Credit Advances Model, 1 April 2025 to 31 March 2026. GOV.UK, 9 July 2026.", "title": "Effectiveness Assessment of Universal Credit Advances Model", "url": "https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model", "date": "2026-07-09", "written_by": "deployer", "quote": null } ], "replication": { "produced_by": [ "A model trained on historical fraud and error data in advance claims. Referrals above a threshold are reviewed by caseworkers alongside control-group referrals with scores removed" ], "must_hold": [ "A human decision on every advance", "Regular fairness assessments (age is the only protected characteristic in the model)", "Feedback of review outcomes into the model" ], "echoes": [ { "date": null, "setting": "DWP fairness assessment of the model for 1 April 2024 to 31 March 2025 (2025–07)", "model_version": null, "outcome": "recurred", "source": "NAO: around three times more effective than a random control sample; over-referral of older and non-UK claimants" }, { "date": null, "setting": "Four further DWP machine learning models in development (2025–26)", "model_version": null, "outcome": "unknown", "source": "NAO: deployment decisions expected in 2025-26" } ] }, "paired_incident": null, "headline": "Universal Credit advances model saves an estimated £4.4 million", "metric": "saving an estimated £4.4 million from 2022–23 to 2024–25 by DWP's estimate", "comparator": { "against": "a randomised control group sample of advance requests", "before": "fraud-risk identification rate of the random control sample (not quantified)", "after": "around three times more effective at identifying fraud risk", "basis": "measured", "quote": "In terms of performance, DWP found the model to be around three times more effective at identifying fraud risk than a randomised control group sample.", "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" }, "later": { "status": "persisted", "date": "2026-07-09", "note": "DWP's effectiveness assessment for April 2025 to March 2026 found the retrained model 2.5 times more effective than a random control group and concluded it should keep operating.", "quote": "the model is 2.5 times more effective at identifying high-risk advances than a random control group, increasing the precision of referrals and enabling human decision-makers to focus on the riskiest cases", "url": "https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model" }, "value": { "amount": "£4.4 million", "what": "advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25", "who": "Department for Work & Pensions", "estimate": true, "quote": "DWP estimates that, from 2022-23 to 2024-25, the model generated savings of £4.4 million that would otherwise have been overpaid.", "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" }, "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": "2026-10-21", "change": "Headline reworded to state the event without naming who claims it.", "source": null, "editor": "AIBD" }, { "date": "2026-10-21", "change": "Title reworded to state the event without naming who claims it.", "source": null, "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Reduced Risk under Trustible's full taxonomy of 23 types; it was Improved Organizational Efficiency 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": "Added what the benefit is measured against, what happened since, the value as stated from the cited sources.", "source": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf; https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model; https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "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://en.wikipedia.org/wiki/Department_for_Work_and_Pensions; https://en.wikipedia.org/wiki/Department_for_Work_and_Pensions", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the replication conditions, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "source": null, "editor": "AIBD" } ], "confidence": "studied", "shade": "high", "claim_status": "Independently studied", "claim_basis": "Sources: one independent report, one independent news report and the deployer's report. The cited evidence includes no finding by a regulator or court.", "profile": { "case_id": "AIBD-0184", "catalog_version": "1.0.0", "source_fingerprint": "537199b2c3da80a94f17562f462dc8ae10d28abb15116260b2664d1adffa355e", "collection_refs": [ { "collection_id": "aibd", "version": "snapshot-8bf7cfec9cb185aa" }, { "collection_id": "aibd", "version": "2026-10-25" }, { "collection_id": "fructus", "version": "2026-10-25" }, { "collection_id": "chanticleer", "version": "2026-10-25" } ], "source_ids": [ "AIBD-0184-source-0001", "AIBD-0184-source-0002", "AIBD-0184-source-0003" ], "families": { "strategic assessments": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] }, "evaluations": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] }, "monetary observations": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "AIBD-0184-legacy-value", "DIV-0031" ], "role": "Existing monetary observations and Fructus statements with matched AIBD case links; each statement retains its own review and eligibility status" }, "provenance": { "kind": "curatorial", "basis": "Linked to the current Fructus dataset by its validated case identifier. Eligible statements can supply unanswered monetary questions without a new source review or a finding of net benefit.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial", "record_ids": [ "AIBD-0184-legacy-value", "DIV-0031" ] }, "institutional decisions": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] } }, "fields": { "actors[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "developer": "Department for Work & Pensions (DWP)", "deployer": "Department for Work & Pensions (DWP)" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "developer", "deployer" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "actors[].ownership_control": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "ai_contribution": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ai_role": { "status": "assessed", "statement": "The NAO describes a model trained on historical fraud and error data that scores advance requests and refers those above a threshold. Every payment decision is made by a caseworker, so the model's contribution is prioritization. Developer and deployer are the same public body (DWP). The NAO, an independent auditor, wrote the report." } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ai_role" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "ai_relationship": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "category": "general benefit", "ai_role": { "status": "assessed", "statement": "The NAO describes a model trained on historical fraud and error data that scores advance requests and refers those above a threshold. Every payment decision is made by a caseworker, so the model's contribution is prioritization. Developer and deployer are the same public body (DWP). The NAO, an independent auditor, wrote the report." } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ai_role", "category" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].alternatives": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].assessor_date": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].depth": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].display_confidence": { "state": "value_present", "value": { "kind": "structured", "data": { "confidence": "high", "evidence_word": "studied", "claim_status": "Independently studied", "basis": "Sources: one independent report, one independent news report and the deployer's report. The cited evidence includes no finding by a regulator or court.", "interpretation": "Existing display rule preserved; the richer outcome-specific confidence question still needs review." } }, "provenance": { "kind": "curatorial", "basis": "Derived from the existing case display properties; not a new evidentiary judgment.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].rationale": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "reason": "The saving is an aggregate departmental estimate over three years, not an attributable outcome." }, { "group": "societal", "reason": "No outcome for claimants or taxpayers beyond the aggregate estimate is documented. Referred claimants bear an added evidence burden." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].rubric_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].status_basis": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].target": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "reason": "The saving is an aggregate departmental estimate over three years, not an attributable outcome." }, { "group": "societal", "reason": "No outcome for claimants or taxpayers beyond the aggregate estimate is documented. Referred claimants bear an added evidence burden." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "availability": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "brief": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "case_kind": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "case_links[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].evidence_links": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].evidence_strength": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].missingness": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].proposition": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "cross_border_flows[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "decisions[].action_outcome": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].grounds": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].institution_task": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].success_criterion": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "deployment_stage": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "economics[].amount": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "value": { "amount": "£4.4 million", "what": "advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25", "who": "Department for Work & Pensions", "estimate": true, "quote": "DWP estimates that, from 2022-23 to 2024-25, the model generated savings of £4.4 million that would otherwise have been overpaid.", "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "value" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].costs[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "economics[].counterfactual": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "DWP’s estimate of payments that otherwise would have been made; method is not detailed in the cited passage.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ] } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].currency_price_basis": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "GBP, identified from the source context. Nominal source amount; price year not separately stated.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ] } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].gross_net": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ] } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].measure_type": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "value": { "amount": "£4.4 million", "what": "advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25", "who": "Department for Work & Pensions", "estimate": true, "quote": "DWP estimates that, from 2022-23 to 2024-25, the model generated savings of £4.4 million that would otherwise have been overpaid.", "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "value" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].period_unit": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "U.K. fiscal years 2022–23 through 2024–25. Aggregate Universal Credit advances model estimate.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ] } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].realization": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "Retrospective estimate.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ] } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].recipient_payer": { "state": "value_present", "value": { "kind": "structured", "data": { "answer": "Recipient: U.K. Department for Work and Pensions and public funds. Payer: DWP benefit expenditure that would otherwise have been overpaid.", "monetary_statements": [ { "id": "DIV-0031", "title": "Estimated Universal Credit overpayments avoided", "monetary_role": "outcome_amount_claim", "financial_type": "avoided_loss", "amount_basis": "avoided_expenditure", "net_benefit_established": false, "source_review_status": "primary_source_checked", "curation_status": "curated", "observations": [ { "id": "primary", "relationship": "primary", "amount": { "source_text": "Estimated £4.4 million", "value": "4400000", "qualifier": "approximately", "currency": "GBP", "currency_basis": "context", "price_basis": "Nominal source amount; price year not separately stated." }, "period": "U.K. fiscal years 2022–23 through 2024–25.", "unit": "Aggregate Universal Credit advances model estimate.", "realization": "retrospective_estimate", "accounting_boundary": "Estimated gross overpayments avoided. DWP could not separate this model’s costs from wider analytics costs.", "source_review_status": "primary_source_checked", "sources": [ { "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "title": "Tackling benefit overpayments due to fraud and error", "quote": "the model generated savings of £4.4 million that would otherwise have been overpaid.", "date": "2025-10-22", "primary": true, "written_by": "independent", "reviewed_at": "2026-10-23", "review_method": "full_text", "locator": "PDF page 31, paragraph 2.11" } ] } ] } ], "original_resolution": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "value": { "amount": "£4.4 million", "what": "advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25", "who": "Department for Work & Pensions", "estimate": true, "quote": "DWP estimates that, from 2022-23 to 2024-25, the model generated savings of £4.4 million that would otherwise have been overpaid.", "url": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "value" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" } } }, "provenance": { "kind": "curatorial", "basis": "Mapped from eligible Fructus statements DIV-0031. Each statement retains its source review status. The mapping performs no new source review or currency conversion and does not establish additional value from AI. The Fructus audit matches the original monetary value. The original field and its provenance are retained with this answer.", "record_fields": [], "source_refs": [ "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf" ] }, "review_state": "record_mapped", "coverage": "partial" }, "economics[].return_payback": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "economics[].valuation_method": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "evaluations[].ai_uplift": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].comparator": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].design": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].effect": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].exposure": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].id": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].measurement_method": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].metric": { "state": "not_assessed", "provenance": { "kind": "legacy_record", "basis": "Mapped from the existing case description without a new external source review.", "record_fields": [ "metric" ], "source_refs": [] }, "review_state": "not_assessed", "coverage": "partial", "family": "evaluations", "display": { "answer": "saving an estimated £4.4 million from 2022–23 to 2024–25 by DWP's estimate", "editor_note": "" } }, "evaluations[].reproduction_materials": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].setting_sample": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].uncertainty_limitations": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "external_references[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "follow_up[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "later": { "status": "persisted", "date": "2026-07-09", "note": "DWP's effectiveness assessment for April 2025 to March 2026 found the retrained model 2.5 times more effective than a random control group and concluded it should keep operating.", "quote": "the model is 2.5 times more effective at identifying high-risk advances than a random control group, increasing the precision of referrals and enabling human decision-makers to focus on the riskiest cases", "url": "https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model" } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "later" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "human_oversight": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "id": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "id": "AIBD-0184" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "id" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "infrastructure_dependencies[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].beneficiaries": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].beneficiary_count": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].benefit_type": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "benefit_type": "Reduced Risk" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "benefit_type" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].claim": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "account": "The National Audit Office (NAO) published “Tackling benefit overpayments due to fraud and error” on 22 October 2025. It states that since May 2022 the Department for Work & Pensions (DWP) has used a machine learning model. The model flags potentially fraudulent claims for Universal Credit advances. DWP estimates it saved £4.4 million from 2022–23 to 2024–25. The model scores new claims and refers those above a threshold to a caseworker. Decisions are made by DWP staff. DWP cannot disaggregate the model's costs from wider spending on data analytics. The NAO reports DWP's finding that the model is around three times more effective at identifying fraud risk than a randomized control group sample. It reports a median payment delay of one day for referred requests. DWP's July 2025 fairness analysis for 2024–25 showed older claimants (age groups 45 to 54 and above) and non-UK nationals over-referred for review. Fairness analysis was possible on only one of nine protected characteristics. Four further models were in development. Civil Service World (Jim Dunton, 22 October 2025) reported the NAO's warnings on data infrastructure. It quoted Transformation Minister Andrew Western saying that the one machine learning model has made the department significantly more effective.", "benefit_type": "Reduced Risk" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "account", "benefit_type" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].distribution": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].duration": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].id": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].stakeholder_group": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "reason": "The saving is an aggregate departmental estimate over three years, not an attributable outcome." }, { "group": "societal", "reason": "No outcome for claimants or taxpayers beyond the aggregate estimate is documented. Referred claimants bear an added evidence burden." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].status": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].tradeoffs": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].unexpected": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "unexpected": false, "purpose": "Assess the fraud risk of Universal Credit advance requests and refer high-risk requests to a caseworker for review." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "unexpected", "purpose" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "places[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "developer_country": "United Kingdom", "deployer_country": "United Kingdom" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "developer_country", "deployer_country" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "purpose": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "purpose": "Assess the fraud risk of Universal Credit advance requests and refer high-risk requests to a caseworker for review." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "purpose" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "record_dates": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "date_observed": "2025-10-22", "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Headline reworded to state the event without naming who claims it.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Title reworded to state the event without naming who claims it.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Reduced Risk under Trustible's full taxonomy of 23 types; it was Improved Organizational Efficiency 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": "Added what the benefit is measured against, what happened since, the value as stated from the cited sources.", "source": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf; https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model; https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "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://en.wikipedia.org/wiki/Department_for_Work_and_Pensions; https://en.wikipedia.org/wiki/Department_for_Work_and_Pensions", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the replication conditions, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "editor": "AIBD" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "date_observed", "revisions" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "replication.attempts[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "A model trained on historical fraud and error data in advance claims. Referrals above a threshold are reviewed by caseworkers alongside control-group referrals with scores removed" ], "must_hold": [ "A human decision on every advance", "Regular fairness assessments (age is the only protected characteristic in the model)", "Feedback of review outcomes into the model" ], "echoes": [ { "setting": "DWP fairness assessment of the model for 1 April 2024 to 31 March 2025 (2025–07)", "outcome": "recurred", "source": "NAO: around three times more effective than a random control sample; over-referral of older and non-UK claimants" }, { "setting": "Four further DWP machine learning models in development (2025–26)", "outcome": "unknown", "source": "NAO: deployment decisions expected in 2025-26" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "replication.conditions": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "A model trained on historical fraud and error data in advance claims. Referrals above a threshold are reviewed by caseworkers alongside control-group referrals with scores removed" ], "must_hold": [ "A human decision on every advance", "Regular fairness assessments (age is the only protected characteristic in the model)", "Feedback of review outcomes into the model" ], "echoes": [ { "setting": "DWP fairness assessment of the model for 1 April 2024 to 31 March 2025 (2025–07)", "outcome": "recurred", "source": "NAO: around three times more effective than a random control sample; over-referral of older and non-UK claimants" }, { "setting": "Four further DWP machine learning models in development (2025–26)", "outcome": "unknown", "source": "NAO: deployment decisions expected in 2025-26" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "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": "2026-10-21", "change": "Headline reworded to state the event without naming who claims it.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Title reworded to state the event without naming who claims it.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Reduced Risk under Trustible's full taxonomy of 23 types; it was Improved Organizational Efficiency 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": "Added what the benefit is measured against, what happened since, the value as stated from the cited sources.", "source": "https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf; https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model; https://www.nao.org.uk/wp-content/uploads/2025/10/Tackling-benefit-overpayments-due-to-fraud-and-error.pdf", "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://en.wikipedia.org/wiki/Department_for_Work_and_Pensions; https://en.wikipedia.org/wiki/Department_for_Work_and_Pensions", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the replication conditions, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "editor": "AIBD" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "revisions" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sector": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "domain": "Benefit fraud and error", "segment": "Universal Credit advances" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "domain", "segment" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].access_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "sources[].citation_locator": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "evidence": [ { "kind": "report", "citation": "National Audit Office, 'Tackling benefit overpayments due to fraud and error', 22 October 2025, paragraphs 15-16 and 3.4-3.15", "title": "Tackling benefit overpayments due to fraud and error", "url": "https://www.nao.org.uk/reports/tackling-benefit-overpayments-due-to-fraud-and-error/", "date": "2025-10-22", "written_by": "independent", "quote": "Since May 2022, DWP has used a machine learning model to flag potentially fraudulent claims for UC advances, saving an estimated £4.4 million." }, { "kind": "news report", "citation": "Jim Dunton, Civil Service World, 22 October 2025", "title": "Digital and data challenges could hamper DWP's fraud strategy, NAO warns", "url": "https://www.civilserviceworld.com/news/article/digital-and-data-challenges-could-hamper-dwps-fraud-strategy-nao-warns", "date": "2025-10-22", "written_by": "independent", "quote": "our one machine learning model has made us significantly more effective" }, { "kind": "report", "citation": "Department for Work & Pensions (2026). Effectiveness Assessment of Universal Credit Advances Model, 1 April 2025 to 31 March 2026. GOV.UK, 9 July 2026.", "title": "Effectiveness Assessment of Universal Credit Advances Model", "url": "https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model", "date": "2026-07-09", "written_by": "deployer" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "evidence" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].id": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "AIBD-0184-source-0001", "AIBD-0184-source-0002", "AIBD-0184-source-0003" ], "role": "Existing evidence items, in their original order" }, "provenance": { "kind": "curatorial", "basis": "Stable within the fingerprinted record version; no source content or source review has been added.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "complete" }, "sources[].language": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "sources[].relationship": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "evidence": [ { "kind": "report", "citation": "National Audit Office, 'Tackling benefit overpayments due to fraud and error', 22 October 2025, paragraphs 15-16 and 3.4-3.15", "title": "Tackling benefit overpayments due to fraud and error", "url": "https://www.nao.org.uk/reports/tackling-benefit-overpayments-due-to-fraud-and-error/", "date": "2025-10-22", "written_by": "independent", "quote": "Since May 2022, DWP has used a machine learning model to flag potentially fraudulent claims for UC advances, saving an estimated £4.4 million." }, { "kind": "news report", "citation": "Jim Dunton, Civil Service World, 22 October 2025", "title": "Digital and data challenges could hamper DWP's fraud strategy, NAO warns", "url": "https://www.civilserviceworld.com/news/article/digital-and-data-challenges-could-hamper-dwps-fraud-strategy-nao-warns", "date": "2025-10-22", "written_by": "independent", "quote": "our one machine learning model has made us significantly more effective" }, { "kind": "report", "citation": "Department for Work & Pensions (2026). Effectiveness Assessment of Universal Credit Advances Model, 1 April 2025 to 31 March 2026. GOV.UK, 9 July 2026.", "title": "Effectiveness Assessment of Universal Credit Advances Model", "url": "https://www.gov.uk/government/publications/effectiveness-assessment-including-statistical-analysis-of-the-universal-credit-advances-machine-learning-model-1-april-2025-to-31-march-2026/effectiveness-assessment-of-universal-credit-advances-model", "date": "2026-07-09", "written_by": "deployer" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "evidence" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].upstream_sources": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "strategic_assessments[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "strategic assessments" }, "summary": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "summary": "The National Audit Office reports that DWP's model, in use since May 2022, flags high-risk advance requests for caseworker review, with estimated savings of £4.4 million and about three times the h...", "account": "The National Audit Office (NAO) published “Tackling benefit overpayments due to fraud and error” on 22 October 2025. It states that since May 2022 the Department for Work & Pensions (DWP) has used a machine learning model. The model flags potentially fraudulent claims for Universal Credit advances. DWP estimates it saved £4.4 million from 2022–23 to 2024–25. The model scores new claims and refers those above a threshold to a caseworker. Decisions are made by DWP staff. DWP cannot disaggregate the model's costs from wider spending on data analytics. The NAO reports DWP's finding that the model is around three times more effective at identifying fraud risk than a randomized control group sample. It reports a median payment delay of one day for referred requests. DWP's July 2025 fairness analysis for 2024–25 showed older claimants (age groups 45 to 54 and above) and non-UK nationals over-referred for review. Fairness analysis was possible on only one of nine protected characteristics. Four further models were in development. Civil Service World (Jim Dunton, 22 October 2025) reported the NAO's warnings on data infrastructure. It quoted Transformation Minister Andrew Western saying that the one machine learning model has made the department significantly more effective." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "summary", "account" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "systems[].autonomy_permissions": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].component_role": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].deployment_configuration": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].inputs_outputs": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].model_name": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].model_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].product_name": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "system": "Universal Credit advances machine learning model" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "system" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "systems[].provider": { "state": "value_present", "value": { "kind": "structured", "data": { "application_developer": "Department for Work & Pensions (DWP)" } }, "provenance": { "kind": "legacy_record", "basis": "Mapped from the existing case description without a new external source review.", "record_fields": [ "developer" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "taxonomy_crosswalks[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "title": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "title": "DWP's Universal Credit advances machine-learning model saved an estimated £4.4 million from 2022–23 to 2024–25", "headline": "Universal Credit advances model saves an estimated £4.4 million" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "title", "headline" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "version": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "537199b2c3da80a94f17562f462dc8ae10d28abb15116260b2664d1adffa355e" ], "role": "SHA-256 of normalized original case data" }, "provenance": { "kind": "curatorial", "basis": "Deterministic content version assigned by the local migration.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "complete" }, "workflow": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "account": "The National Audit Office (NAO) published “Tackling benefit overpayments due to fraud and error” on 22 October 2025. It states that since May 2022 the Department for Work & Pensions (DWP) has used a machine learning model. The model flags potentially fraudulent claims for Universal Credit advances. DWP estimates it saved £4.4 million from 2022–23 to 2024–25. The model scores new claims and refers those above a threshold to a caseworker. Decisions are made by DWP staff. DWP cannot disaggregate the model's costs from wider spending on data analytics. The NAO reports DWP's finding that the model is around three times more effective at identifying fraud risk than a randomized control group sample. It reports a median payment delay of one day for referred requests. DWP's July 2025 fairness analysis for 2024–25 showed older claimants (age groups 45 to 54 and above) and non-UK nationals over-referred for review. Fairness analysis was possible on only one of nine protected characteristics. Four further models were in development. Civil Service World (Jim Dunton, 22 October 2025) reported the NAO's warnings on data infrastructure. It quoted Transformation Minister Andrew Western saying that the one machine learning model has made the department significantly more effective.", "replication": { "produced_by": [ "A model trained on historical fraud and error data in advance claims. Referrals above a threshold are reviewed by caseworkers alongside control-group referrals with scores removed" ], "must_hold": [ "A human decision on every advance", "Regular fairness assessments (age is the only protected characteristic in the model)", "Feedback of review outcomes into the model" ], "echoes": [ { "setting": "DWP fairness assessment of the model for 1 April 2024 to 31 March 2025 (2025–07)", "outcome": "recurred", "source": "NAO: around three times more effective than a random control sample; over-referral of older and non-UK claimants" }, { "setting": "Four further DWP machine learning models in development (2025–26)", "outcome": "unknown", "source": "NAO: deployment decisions expected in 2025-26" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "account", "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" } }, "summary": { "total": 91, "addressed": 91, "known": 36, "reviewed": 0, "record_mapped": 36, "not_assessed": 55, "complete": false, "by_state": { "value_present": 36, "not_reported": 0, "unresolved": 0, "not_assessed": 55, "not_applicable": 0, "withheld": 0, "source_unavailable": 0 } } } }