I Reduced Task Time
Reducing the time it takes an individual to perform a discrete task.
Depth describes the extent of a benefit, from convenience to a life or livelihood preserved. Evidence is assessed separately.
Full search could not load. Results below match titles only.
Each segment opens its cases. Years follow the dates shown in the records, including publication-based dates.
Reducing the time it takes an individual to perform a discrete task.
Reducing the financial costs to an individual to perform a discrete task.
Customization or personalization of a product/service that increases usability or user value.
Increased access to new ideas, or innovative ways of creating products/services.
Improved physical health outcomes or information about bodily health.
Improved educational outcomes or experiences including faster learning rates or academic achievements.
Improved safety features, detection mechanisms, or failsafes.
Enabling a process to efficiently scale to additional users, sectors, or business units.
Increased accuracy or performance for a discrete task.
Easier and more accurate access to information necessary to accomplish a task.
Increased probability that a customer continues to use a product/service.
Automating highly inefficient repetitive tasks, particularly those that contribute to harmful bottlenecks or bloated bureaucracy.
Reduction in potential for harm, especially potential legal harm.
Improvements to overall operational efficiency, including helping facilitate cross team processes, reducing organizational friction, or reducing redundancies.
Direct increases in product/service revenues or contract growth rates over time.
Improved social standing and reputation with core stakeholders.
Improved diversity of communities or groups of people.
Increased acceptance of people with diverse backgrounds.
Improved environment including lowered carbon emissions, reduced waste, or using recycled materials.
Increased accessibility in society for people with disabilities.
Reduction in energy usage, water consumption, or direct carbon emission outputs.
Improved access to educational services, institutions, or programs.
Reduction in criminal activities or physically harmful incidents.
| Compare |
|---|
Nothing selected yet. The Compare control beside each case adds it here. Selection cleared.
Read the cases the way your work needs them. A role adds its own line to every case and orders the list for it.
Observed
Anthropic flag on a shooting threat leads to a Florida arrestArrest report and probable cause statement, LCSO report 26-530248 (Heller, Carli Michelle), PublishedObserved
Waymo pulls a car over; police find a loaded ghost gun with two minorsWaymo pulls over, calls police on juvenile riders who had 'ghost gun,' authorities say, PublishedObserved
Drivers slowed down when signs warned of wildlife aheadElk crossing: Real-time driver warning system helps prevent animal collisions, PublishedInformation as of
Researchers found 47 known gravitational-wave signalsMADGRAV: a multilevel anomaly-detection pipeline for gravitational-wave searches applied to LIGO data, PublishedObserved
Police facial recognition vans in Huddersfield lead to three arrestsCutting Edge Facial Recognition Cameras Used for First Time in Kirklees, PublishedObserved
Odakyu trains stop when software detects a person left in a crossing踏切に取り残された人をAIで検知→列車を自動停止 小田急が実運用, PublishedObserved
Drone software detects two lost hikers, who are found within five hoursFire and Rescue NSW drone team locates missing hikers using AI technology, PublishedObserved
OpenAI flag leads Brazilian police to hold a father a day earlyAlerta do FBI e do Ministério da Justiça auxiliam PCES na prisão de investigado por ameaças contra o próprio filho em Inteligência Artificial, PublishedObserved
AI cameras identify 18 teenagers gathering with weapons in Việt TrìTừ camera AI, công an chỉ mất 30 phút nhận diện 18 thanh thiếu niên tụ tập đánh nhau trên phố, PublishedObserved
Saronic's Corsair sea drone picks up two Apache crew members off OmanU.S. Army Crew Safely Rescued After Helicopter Lost at Sea, PublishedObserved
Meta's suicide alert brings Meerut police to a man within 8 minutesWhen every minute counts: Uttar Pradesh Police save youth attempting suicide in 8 minutes after Meta alert over Instagram post, PublishedInformation as of
IMIE intelligent retina restores partial vision to a blind woman全国首例!脑机接口智能视网膜让盲人看见字符, PublishedObserved
Qualcomm's humanoid may have collapsed safely at ComputexHumanoid Robot Malfunctions and Collapses During Live AI Demo at Computex 2026, PublishedObserved
OpenAI refers Darren Zhou's threats to the FBI; he pleads guiltyMan Flagged by ChatGPT Pleads Guilty in Stalking Case, PublishedThe April 6 White House briefing documents the rescue of a downed American airman in Iran.[1] The claim that an AI-assisted quantum sensor located him comes from two anonymous sources in the New York Post. President Trump subsequently told the Post that the tool was important, without explaining how it worked.[2] Physicists interviewed by the Post and Scientific American questioned the described capability.[3] The rescue is established; the contribution attributed to Ghost Murmur remains unverified.
Suggestions and corrections receive a public tracking number on GitHub. A GitHub account is required. Email remains available for private material.
Public submissions and their status. Peter H. Vartanian reviews contributions before publication.
... which is why this database exists: to help others benefit from AI that has already helped someone else.
Edited by Peter H. Vartanian.
An AI benefit is a good that a person, an organization or society has received, or is intended to receive, to which an identifiable AI system contributed. An AI-related topic alone is insufficient. The editor’s guide sets out how cases are written and graded. Reported outcomes and intended benefits have distinct labels. Every case has a beneficiary shape, a depth color, and a confidence fill. Gray means that depth is unassessed.
A record describes one of four kinds of case: a benefit created or restored, a specific harm prevented, a protective response whose prevention is unproven, or a general benefit case. General cases can include launches, simulations, accuracy studies, aggregate claims, cohorts, AI-assisted learning, and proposed applications. Their category does not determine their depth or confidence.
Each beneficiary group that benefited is rated from 1 to 5, the depth of the benefit for that group: 1, a convenience, or time or cost saved, with life otherwise unchanged; 2, a measurable improvement in a result or a capability, such as a test score or a crop yield; 3, a lasting change, a capability restored or a risk removed; 4, a serious harm averted or a serious condition resolved in time; 5, a life or a livelihood preserved, or a change documented across a population. Every rating has a reason. Depth describes the magnitude of the benefit that the source reports; it does not establish that the claim is independently verified. Aggregate outcomes and gains in capability can be rated. Intended benefits remain explicit, and a group stays gray when the extent of its benefit cannot be established. Corrections appear in the revision history.
What the system contributed to the benefit is assessed separately from whether the benefit occurred, and may be left unrated. A system that was present is not thereby responsible.
A system’s name, base model, adapted variant, interface, hosting, and settings are assessed separately. A supported inference appears with its basis in an editor’s note. Later documentation may identify a plausible model, but does not establish that it was available or used on the event date. The importance of a task does not establish its reasoning settings.
Source review traces each finding to the original work that produced it. Different claims in one case may have different primary sources. An additional source must contribute evidence, verification, a correction, or a material limitation. Repetition does not constitute corroboration. A primary source may have an interest in the result, so origin and independence are assessed separately. When the original cannot be inspected, the record identifies the intermediary and the limitation.
Each source identifies its kind and its author's relationship to the system: maker, deployer, beneficiary, evaluator, independent author, or unassessed. An evaluator's role alone establishes neither independence nor a conflict of interest.
Only sources that support the benefit determine its evidence grade. Among independent sources, a court or regulatory finding gives the grade found. A peer-reviewed study, clinical case report, or research report gives studied. A descriptive report, news report, preprint, or dataset gives reported. The strongest qualifying source determines the grade. Without one of these sources, the grade is claimed. This includes accounts from interested parties, even when published as studies, and sources whose independence is unassessed.
The fill is solid for found and studied, hatched for reported, and outlined for claimed. These correspond to high, medium, and low confidence. Contrary evidence, procedural documents, and background material remain in the sources without raising the benefit's grade.
Source notes describe authorship, independence, and any documented conflicts of interest. They identify gaps in the cited evidence. Inferences are labeled.
Briefs by Peter H. Vartanian examine the evidence and its gaps. Their dates appear in the revision history.
Each entry includes fields for the conditions that produced the benefit and the conditions that would have to hold again. Subsequent attempts appear with their setting, model version when reported, and outcome. An outcome that the cited sources do not establish remains unresolved.
Where an incident database holds an entry about the same kind of system harming the same kind of stakeholder, the record links to it, so that a reader weighing benefit against risk can open both. No severity is shown for the harm.
Cases can be filtered and viewed as a list, table, map, timeline, or table of benefit types. The view and the filters are kept in the address, so that a view can be shared.
The review log gives the reasons for excluding a candidate. Unfavorable results remain visible when they bear on a reported benefit. Multiple sources about the same benefit belong to one entry. Intended benefits, aggregate results, and weak evidence can qualify.
A case distinguishes an observed event date from the publication date, the date through which a source says that its information is current, or the date on which the source was checked. Month-only dates display the month without an invented day. The timeline uses that date only when an event or source supports it. Dates of access and revision do not determine the order.
Newest follows event or source dates. When a case's date basis has not been assessed, a dated source must support the same date before the case appears in Newest. That date is labeled as a source publication date. Source-check and revision dates do not determine the order. Cases without a supported date remain in the full database.
Counts on this site describe the database's coverage. They do not measure how beneficial AI is, and a trend in records is a trend in reporting.
The map in the Gazetteer draws Natural Earth's 1:110m countries with de facto boundaries in an equirectangular projection. A country's circle has an area in proportion to its cases; a route runs from the country where a system was made to the country where the case happened. Cases that state no country are not on the map.
Excluded candidates and their reasons. An unresolved AI contribution can warrant further evidence. Intended benefits, aggregate results, and weak evidence can qualify.
These systems were not searched again because the database already records them. Face2Gene is № 0037, the Apple Watch and other wearables are № 0021 to № 0027, the speech neuroprostheses are № 0002 to № 0005, and the brain-spine interface is № 0001. A new entry would repeat those records.
This OECD AI Monitor entry covers the press coverage of June 2026 (Washington Post, 15 June; MIT Technology Review, 16 June) on the BrainGate2 speech neuroprosthesis from UC Davis that Casey Harrell uses. That case is already № 0002, and its record cites the NEJM paper of 2024 and the Nature Medicine paper of 2026 on long-term use.
This OECD AI Monitor entry describes the same event as № 0010. Meta alerted the Social Media Center of the UP Police at 6:53 pm on 7 June 2026, and Sardhana police reached the home of the 25-year-old within eight minutes. The entry draws on PTI reports carried by Rediff, ETV Bharat, The Hans India and LatestLY.
This OECD AI Monitor entry describes the same event as № 0062: Mia flagged a 6 mm cancer that NHS Grampian radiologists had missed during the GEMINI evaluation. The entry draws on the BBC report of 21 March 2024 and a University of Aberdeen release. The Aberdeen release of 10 March 2026 has since been added to the evidence for № 0062.
This OECD AI Monitor entry describes the same arrests of 39 poachers in Similipal as № 0102. Its sources are The New Indian Express, the Times of India and PTI, dated 20 and 21 February 2026. These reports date the entry of the poachers to the night of Thursday, 19 February. The account of the Field Director gives the night of Wednesday, 18 February.
Ten OECD AI Monitor entries on elephant detection by Indian Railways, dated from March 2023 to February 2026, repeat two cases that the database already records. These are the intrusion-detection case in Assam, № 0009, and the thermal-camera case at Madukkarai, № 0101.
This OECD AI Monitor entry, titled "Tesla Autopilot Prevents and Fails to Prevent Collisions: Contrasting Incidents," matches a case that is already recorded as № 0109. That case concerns a Tesla Model S and a wrong-way driver in Cabarrus County.
This OECD AI Monitor entry, titled "Tesla Autopilot Prevents Fatalities During Storm Dennis by Emergency Braking," is the case of a Tesla Model X and a falling oak tree. That case is already recorded as № 0108.
This OECD AI Monitor entry combines two events: the crash into an overturned truck in Taiwan, which is a harm listed as incident 218 in the AI Incident Database, and a video of a car avoiding a "pig." Reports in Electrek (8 June 2020) and the Daily Star (9 June 2020) show that the video is the swerve of Kenneth Van Hoesen around a wild boar in Belgium on 30 May 2020. That event is written up from a separate Monitor entry and has no case number yet.
№ 0233 already records Operación Influencer (11 May 2023), the Fiscal Heredia model and the ten detentions. That record also cites this OECD AI Monitor entry.
This is the same drive as № 0139: Max Paul Franklin's Tesla Model Y on FSD v12 took him to a VA emergency room on 2 April 2024. In the Bosnian and Croatian reports behind this entry (Raport.ba, 16 April 2024, citing a post by Elon Musk of 11 April), the story becomes a 20 km drive in Virginia during a heart attack. The record for № 0139 already notes these garbled syndicated versions.
№ 0232 already records the facial recognition match by Cheshire Constabulary to Keenan Hughes, the stop with Vincent Potter on 17 May 2024 and the sentences of 1 July 2024. That record also cites this OECD AI Monitor entry.
№ 0234 already records the camera-analysis system by Asilla at about 40 stations and buildings in Japan and the two suicide interventions that Kyodo reported on 1 June 2026. This OECD AI Monitor entry covers the same event with the same sources: Kyodo, the Straits Times, Anadolu and syndicated copies.
№ 0263 already records the Neuralink N1 implant that Noland Arbaugh received in January 2024 and his control of a cursor and a chess game by thought. This OECD AI Monitor entry covers the same person, system and event. The Monitor labels the entry an incident because of its material on malfunction and privacy, and that material describes no separate benefit.
The authors of the Nature Communications paper state that they did not implement a multichannel decoding algorithm. The patient selected letters through firing-rate thresholds with auditory neurofeedback. The only support-vector machine belonged to an earlier eye-tracking system, so no machine-learning system produced the benefit.
ChatGPT served only to draft and edit the Cureus paper about a delayed diagnosis of pneumothorax at Jersey General Hospital. The AI content of the paper is a proposal of how AI could have helped. No AI system had a role in the care of the patient.
Raleigh's Lights Out (27 August 2026) follows a fixed seasonal schedule from 11 p.m. to 6 a.m. and lists BirdCast only as a public resource. The Bird Safe Philly page has no mention of BirdCast alerts and no collision results, so no AI output changed either decision. Houston's Lights Out Nights is the case that BirdCast drove, and it is № 0103.
The only event in this OECD AI Monitor entry is a Titian robot ablation on a 64-year-old woman with paroxysmal atrial fibrillation at Shaoxing People's Hospital, reported by Xinhua Zhejiang on 13 December 2025. The surgeon steered the teleoperated catheter robot from a console, and the report has no mention of AI, autonomy or inference. The CA-GPT item in the entry is a comparative preprint with 96 patients and reports no event. The third source concerns an unrelated neurosurgical robot in Fushun.
Independent studies exist, including a 2026 trial with 11 participants at Nimes University Hospital in the Journal of NeuroEngineering and Rehabilitation. The exoskeleton walks on trajectories that nonlinear optimization computes in advance, and a hybrid zero dynamics controller tracks them. No machine learning or other AI is involved.
The Estonian Ministry of Justice stated on 16 February 2022 that it does not develop an AI robot judge to replace the human judge in small-claims or general court proceedings. There is no system and no outcome to verify.
Human bank staff made the intervention in the Rajesh Dandotiya case. No AI system was involved, so there is nothing to record.
The Singapore Police Force says that Project ASTRO uses robotic process automation to send SMS alerts to scam victims (MHA Minister's Awards page, 29 October 2024). The Straits Times reported on 16 April 2023 that officers and OCBC staff analyzed fund flows and used the tool to send SMS messages in bulk. The S$442 million that police say has been averted since April 2023 is a total for automated messaging and does not measure model-based detection.
The Halo is a titanium bar over the cockpit that the FIA made compulsory from 2018. Neither Wikipedia nor the DER STANDARD report of 13 September 2021, in which Wolff says "Der Halo hat definitiv Lewis' Leben gerettet," has any mention of a sensor, software or an AI component. The sources that the Monitor lists are match reports that call the Halo a safety system. The label "AI-enabled" comes from the Monitor alone.
No peer-reviewed case report was found in which AI embryo selection or AI newborn screening changed the care of an identifiable patient. There was no system or case to examine.
No individual case of a patient helped by a sepsis early-warning alert was found in a source independent of the developer. Searches returned only cohort studies and implementation stories. The cohort evidence for TREWS is № 0066.
ChatGPT-3 and ChatGPT-4 were run on the Milan case only after the diagnosis had been made. Neither model suggested the correct diagnosis, and neither had any part in the care of the patient. The result is a failed test of the models, so there is no benefit to record.
No peer-reviewed case report was found of a Garmin alert that led to a diagnosis. Bogár et al. (2024) searched for Garmin cases and found none, and case-report searches in Europe PMC returned no qualifying case.
No peer-reviewed case report was found of a WHOOP alert that led to a benefit. Searches returned only validation studies and vendor blog posts.
Searches returned only studies that reclassified the CHIME/FRB catalogs with machine learning. The searches found no AI-detected burst at CHIME in 2024 with independent confirmation, so there is no system or event to record. The Berkeley neural network that detected pulses from FRB 121102 is № 0018.
No statement in which a museum accepts an AI-based attribution was located, and the candidate included no named system, painting or museum. The other half of the candidate, Ithaca, is № 0017. The one named attribution case, the de Brécy Tondo in Bradford, came up under another theme and is № 0276.
No peer-reviewed acceptance was found of any AI identification of the language of the Voynich manuscript. No system, source or case was examined.
No machine-learning attribution of a work to Bach was found that is peer-reviewed and accepted by scholars. No system, source or case was examined.
No deployed system for sign-language recognition or translation with an independently evaluated outcome for deaf users was found. The authors of a 2026 systematic review in JMIR on such systems for health-care communication with deaf people report that every included system was at the stage of development and testing. Authors whom they contacted confirmed that none had been deployed or commercialized.
Searches returned only aggregate statistics from industry and regulators on authorized push payment fraud. No court or regulator record was found of a specific AI interception with a named victim, and the candidate included no named system, victim or event. The cases of AI fraud interception now in the database, such as the TIAA case (№ 0210), came from other candidates.
No regulator, court or peer-reviewed record was found of an individual intervention by automatic emergency braking in which the crash did not occur. The NTSB documents that were located, on Uber in Tempe in 2018 and Ford BlueCruise in 2026, concern failures or disabled braking. The aggregate IIHS study is № 0008, and the Tesla incidents recorded from press reports (№ 0108 and № 0109) came from other candidates.
No UIC document naming an AI detection system or a trial result could be located. The safety pages on uic.org list the TreSP network, the TRESPAD events and project names such as AI4SAFEBEHAVE with no results, and the UIC page on trespass and suicide returned a 404 error. The only identifiable deployment is the one by ProRail, which is recorded separately as № 0309.
No statement was found in which an authority documents an evacuation or emergency decision taken on an AI flood or nowcast forecast. The candidate included no named system, place or event of its own, and the cases that it pointed to are already № 0031 (flood alerts in Bihar) and № 0092 (skill evaluation in Mali). Records of action on AI forecasts from other candidates include the anticipatory cash payments in Kogi (№ 0095) and the activation by the Philippine Red Cross (№ 0098).
SyRI is a documented harm. The District Court of The Hague invalidated the system in 2020, so it was set aside and was not examined as a benefit.
In the procurement assessment of the council and the service documents of Huume, the housing improvements come from configurable workflows, policy calculations, portals and dashboards. None of the reviewed sources ties an identified AI function to those gains. Dorset's separate pilots of meeting summaries and transcription do not show that AI contributed to this housing case.
The dataset is one file, records.json, released under CC BY 4.0. Three flat tables are derived from it: records.csv, one row per record; ratings.csv, one row per beneficiary group rated; and echoes.csv, one row per attempt to make a benefit happen again. The schema describes every field. Each record is also available on its own page as JSON.
Dataset editions are archived at Zenodo, each with its own DOI and a manifest of file checksums. Each record’s citation names its version and the edition that holds it.
Measured outcomes: CSV or JSON. Taxonomy mappings: CSV or JSON. The research schema defines these tables.
Field definitions and symbols.
Download the public case profiles, field dictionary, and collection profiles. The coverage table lists each public field and its status for every case.
AI outcomes, in money. A separately curated collection of financial claims, with original currencies, periods, recipients, comparisons, and supporting sources. JSON and CSV.
One system, one setting, one benefit is one record. A benefit reported in several sources is one record with several sources. A benefit that happened again elsewhere is an echo on the original record, not a new record.
Artificial Intelligence Benefits Database is made by Peter H. Vartanian. Write to peter.vartanian@icloud.com.
The dataset and this site are released under CC BY 4.0. Benefit types and beneficiary groups follow Trustible's benefits taxonomy.
The border adapts the color and grayscale reference strip shown beside a plate from E. A. Séguy’s Papillons in Smithsonian Libraries’ digitization.
Set in Newsreader, IBM Plex Sans and IBM Plex Mono, all used under the SIL Open Font License. Sector icons are Bootstrap Icons, under the MIT License. Organization logos are the organizations' own marks, shown to identify them; each one's source is listed in logos.json. Built 29 October 2026.
In U.S. police-reported crash data, forward collision warning with automatic emergency braking cut front-to-rear crashes by 49%, and automatic braking cut pickup rear-end crash rates by 43%.
front-to-rear crashes were reduced by 49%
In RECOVERY, 12% of patients receiving baricitinib died within 28 days, compared with 14% receiving usual care alone. [2]
In ACTT-2, patients receiving remdesivir recovered in a median of 7 days with baricitinib, compared with 8 days with a placebo. [5]
In an Insurance Institute for Highway Safety study, forward collision warning alone, low-speed emergency braking and warning with braking cut rear-end striking crash rates by 27%, 43% and 50%.
reduced rear-end striking crash involvement rates by 27%, 43% and 50%
In Australian and New Zealand injury crashes from 2016 to 2023, cars with pedestrian-detecting AEB had 17% fewer pedestrian or cyclist collisions than cars without AEB.
a 17% lower rate of pedestrian or cyclist collisions (95% CI 5% to 27%)
Fall Detection on an Apple Watch Series 10 called emergency services and alerted emergency contacts when a 69-year-old runner collapsed in cardiac arrest, and he made a full neurological recovery.
began cardiopulmonary resuscitation about 6–7 minutes after the collapse
An IIHS study of US police-reported crashes found that automatic emergency braking with pedestrian detection cut pedestrian crash risk by 25% to 27% and injury crash risk by 29% to 30%.
a 25% to 27% lower pedestrian crash risk and a 29% to 30% lower pedestrian injury crash risk
In a Taiwan trial of 15,965 patients, a neural network flagged high-risk electrocardiograms for physicians, and 90-day mortality was 3.6% versus 4.3% with usual care.
3.6% versus 4.3%
An Embrace wristband detected a seizure that left Morgan McGrath, 26, of Fulton, Missouri, not breathing and alerted her mother, who called paramedics; IEEE Spectrum reports that she survived.
Two anonymous sources credit the CIA's Ghost Murmur, quantum sensing paired with AI, with locating a downed F-15E officer in Iran; CBS reports that CIA technology pinpointed him and names no tool.
one rescued airman; the claimed AI contribution remains unverified
On 22 February 2026 near Batavia, Ohio, a 19-year-old's truck flipped into the Little Miami River; Clermont County officials say that her Apple Watch Crash Detection called 911, and crews freed her.
Lynxight, a camera system at a Perth pool, alerted a lifeguard's smartwatch to a man who had sunk unconscious; lifeguards and a swimmer rescued him, and he recovered fully.
Software, apparently Viz.ai, detected Julie Knichel’s brain bleed and alerted her doctors; after treatment, she returned to full-time nursing.
Every Cure’s early drug-repurposing system gave high scores to three drugs that a consulting physician had considered for Joseph Coates’s POEMS syndrome; he responded and later reached remission.
he responded within one or two weeks
After UC San Diego Health deployed COMPOSER, a deep-learning model that alerts nurses to sepsis, mortality was 9.49%, against a modeled 11.39% without it.
1.9 percentage points
Mayo Clinic says that AI algorithms, which no source names, let its care team find Sophia Chan's ruptured aneurysm in minutes; she had surgery and recovered.
In Japanese police data for 2016 to 2019, AEB on six best-selling car models was linked to 20% lower odds of fatal or serious injury for struck pedestrians; cyclists showed no clear change.
20% lower odds of a fatal or serious injury (adjusted odds ratio 0.80, 95% CI 0.64 to 0.996)
A smartwatch ECG function of an unnamed brand recorded a 165 bpm wide-complex tachycardia during palpitations, and cardiologists diagnosed ventricular tachycardia and implanted a defibrillator.
recorded a 165 bpm wide-complex tachycardia during palpitations
Across 463,094 women at 12 German screening sites, radiologists who used Vara MG found 6.7 cancers per 1,000 against 5.7 without it, with no rise in recalls.
found 6.7 cancers per 1,000 against 5.7 without it
In the MASAI trial of 105,934 women, Transpara-supported screening found 6.4 cancers per 1,000 against 5.0 with standard double reading and cut screen reading by 44.2%.
a detection rate of 6.4 per 1,000 against 5.0 per 1,000
At Capio Sankt Göran Hospital, Lunit's Insight MMG read 55,581 mammograms beside two radiologists; the three together found 269 cancers against 250 by the radiologists alone.
the three together found 269 cancers against 250 by the radiologists alone
The Fitbit Alta HR app showed a woman in Zurich that her resting heart rate had risen from about 55 to about 75 beats per minute, and the trend led to an early diagnosis of large vessel vasculitis.
resting heart rate had risen persistently from about 55 to about 75 beats per minute
Orthopedic surgeons at Toronto Western Hospital report that a patient's smart wristwatch, of a brand that the abstract does not name, prompted readmission for a postoperative pulmonary embolism.
In an English National Health Service pilot, Exomiser ranked candidate variants for 2,183 families; 25% of probands got a genetic diagnosis, and a quarter of diagnoses were immediately actionable.
Genetic diagnoses were found in 25% of probands, rising to 35% for monogenic disorders
Meta's suicide alert about a 25-year-old man's Instagram video brought Meerut police to his home within eight minutes, and his condition improved after hospital treatment.
a Sardhana team reached his home within eight minutes
Meta's suicide detection and review system flagged an 18-year-old's Instagram post; Raebareli police reached him within 12 minutes, and he was treated after taking rat poison.
A police team reached his home within 12 minutes
Hanoi police traced a lost 78-year-old on AI traffic cameras on 15 December 2025; a passer-by found him at 3 a.m. the next day and took him home.
On 16 October 2024 Northeast Frontier Railway's AI-based Intrusion Detection System alerted the crew of the Kamrup Express, who also saw about 60 elephants crossing in Assam and braked.
saved 414 elephants in 2023 and 383 up to 16 October 2024
Tesla Full Self-Driving (Supervised) v12 drove Max Paul Franklin 13 miles to a VA emergency room at 2 a.m. on 2 April 2024, according to his own uncorroborated post on X.
drove Max Paul Franklin 13 miles to a VA emergency room
U.S.-led investigators credited facial recognition, reported as Clearview AI, and AI image enhancement for leads in a 2023 operation that made 311 probable identifications of child abuse victims.
311 probable identifications of previously unknown victims, 14 positive contacts
Vyntra's “Fraud Cases” page credits its AI risk models with blocking payments of USD 7,500 to CHF 1,000,000 and a USD 500,000 (KSh 58 million) ATM fraud that ended in arrests.
fraudsters were arrested while attempting to withdraw USD 500,000 (KSh 58 million)
The Turkish police's ASENA analysis program flagged a car at Yüksekova on 20 March 2022, and officers found 52.5 kg of explosives that the interior minister said were bound for Istanbul.
133 packets, 52.5 kg, of factory-made explosive
WeChat says that its WeChat Pay anti-fraud reminder connected a Yunnan man to customer service, which advised him to cancel a 190,000 yuan transfer to a “goddess” he had met online.
avoided a loss of 190,000 yuan
SARUAV, software that uses convolutional neural networks, detected an unconscious 65-year-old man in tall grass on drone images after more than 24 hours, and mountain rescuers retrieved him alive.
After 4 h 31 min of analysis the system detected him and returned his coordinates
In April 2019 Thorn's Spotlight, using Amazon facial recognition, matched a missing child's photo to online sex ads; Thorn's chief executive said that she was recovered within weeks.
the child was recovered within weeks
A study by the Veterans Affairs office that runs REACH VET found that veterans flagged by its suicide-risk model had fewer suicide attempts and admissions and no change in deaths.
documented suicide attempts (-0.05; -0.06 to -0.03)
On 14 January 2026 the thynC monitoring system flagged ventricular tachycardia in an 85-year-old on a general ward in Gwangju, and staff resuscitated the patient.
heart rate climbed to 192 beats per minute
Justice Department analysts detected anomalous Medicare billing through proactive data analytics, and $4.41 billion was held in escrow; no source names the software or says that it used AI.
held $4.41 billion in escrow from about $4.45 billion scheduled for payment
XRlabs says that its Surgical Intelligence platform designed skull implants for conjoined twins, and the hospital credits AI planning in their successful separation in four staged procedures.
successful separation in four staged procedures
Kent County Council reports that under 1% of households flagged by a Xantura predictive-analytics platform and given proactive support became homeless, against 40% of an unsupported group.
Alibaba DAMO's PANDA model exceeded mean radiologist sensitivity for pancreatic cancer on CT scans by 34.1% and found five cancers that the initial standard of care had missed.
China Medical University Hospital’s Intelligent Anti-Microbial System flagged ceftazidime resistance in 1 hour, and doctors changed a 78-year-old woman’s antibiotic before she recovered.
flagged ceftazidime resistance in 1 hour
Takeda reports 36.5% complete skin clearance with zasocitinib versus 13.9% with deucravacitinib at week 16 for a drug that Nimbus designed with Schrödinger's computational platform.
36.5% complete skin clearance with zasocitinib versus 13.9% with deucravacitinib at week 16
A smartphone app that the authors describe as powered by Google's Med-Gemini-3D suggested a rare skull finding that led to a genetic diagnosis in a 22-month-old; the app is unidentified.
A patient's dialogue with DeepSeek-R1-0528 pointed to familial Mediterranean fever, genetic testing confirmed it, and colchicine stopped attacks after ~30 years of symptoms.
colchicine stopped attacks after ~30 years of symptoms
In a two-hospital Seoul trial of 998 patients, colonoscopy with Ainex's ENAD-DET-01 found adenomas in 52.3% of patients against 36.1% with standard colonoscopy.
found adenomas in 52.3% of patients against 36.1% with standard colonoscopy
Columbia University's STAR system found seven sperm, two of them motile, in 2.5 million images of a “sperm-free” sample; a couple who had tried for 19 years then had a girl.
found seven sperm, two of them motile, in 2.5 million images of a “sperm-free” sample
A smartwatch of unstated brand flagged arrhythmias that led to a diagnosis of cardiomyopathy in a 34-year-old man; radiofrequency ablation restored his ejection fraction to 50% at six months.
radiofrequency ablation restored his ejection fraction to 50% at six months
A Withings ScanWatch displayed a prolonged QT interval that led clinicians in Germany to diagnose inherited long QT syndrome in a father and his daughter and to prescribe nadolol.
displayed values such as QT 506 ms and QTc 457 ms
A 76-year-old saw a transient negative T-wave in a recording from his Apple Watch ECG app after exercise; tests showed three-vessel coronary disease, and he had four bypass grafts.
he received four bypass grafts; recovery was uneventful
Over 3,868,506 rider-only miles, the Waymo Driver had no bodily-injury claims and 76% fewer property-damage claims than a Swiss Re human baseline; a later independent study found 68% fewer crashes.
0.78 property-damage claims per million miles against 3.26, a 76% reduction
At 42 VA facilities chosen at random to receive GI Genius, adenoma detection rose from 50.7% to 54.9%, while it fell from 51.8% to 51.1% at 97 control facilities.
adenoma detection rose from 50.7% to 54.9%
The Blood Oxygen app on an Apple Watch Series 7 recorded an oxygen saturation of 88%, which led clinicians in Japan to computed tomography angiography that found a pulmonary embolism.
Apple Watch Series 7 had shown an oxygen saturation of 88% in the early morning
A smartwatch of unstated brand alerted a 58-year-old man to an irregular heart rate above 180 beats per minute; the care that followed found a lung adenocarcinoma, which surgeons removed.
The 35 mm tumor was staged pT2a pN0 M0 with clear margins
A Fitbit Sense watch detected an irregular heart rate of 164 beats per minute in a man in Jamshedpur, India, who sought care and was treated for ventricular tachycardia.
detected an irregular heart rate of 164 beats per minute with irregular rhythm
A Synchron Stentrode implant with a participant-specific decoder turned attempted movements into clicks, and with eye tracking it let two people with ALS use a computer at home.
typed 13.81 and 20.10 correct characters per minute
In six California memory care facilities, real-time alerts from SafelyYou Guardian's AI video fall detection cut residents' time on the ground after a fall by 29.6 minutes.
time on the ground fell by 29.6 minutes (95% CI 20.3 to 38.9)
Mr. Joseph Holycross said that an emergency room doctor confirmed the atrial fibrillation that the Samsung Health Monitor electrocardiogram app detected on his Galaxy Watch4 Classic.
About ten days later he returned for open-heart surgery
On a Fire and Rescue NSW drone, AI software from a vendor that no source names flagged two lost hikers in Kosciuszko National Park, and they were located within five hours.
located less than five hours after the report
OpenAI's safety systems flagged a ChatGPT user's plan to kill his 8-year-old son; alerted through the FBI, Brazilian police held him a day before his stated date.
held him a day before his stated date
A remotely piloted Saronic Corsair sea drone carried two Apache crew off Oman to a helicopter hoist; Central Command reported that both were rescued within approximately two hours.
both were rescued within approximately two hours
After Xiangya Hospital implanted the IMIE intelligent retina, a 61-year-old woman blind for over ten years could identify E optotypes and walk indoors without an escort.
she could identify E optotypes (acuity 0.03, peak 0.1)
Asilla's behavior-analysis AI for security cameras alerted guards, who then reached two people at apparent risk of suicide, the company says; the facilities and dates are undisclosed.
two people at apparent risk of suicide
Through the FBI, OpenAI disclosed threats that Darren Zhou wrote in ChatGPT against his ex-girlfriend, and he later pleaded guilty to stalking and threat charges.
eight years of probation, two years of GPS monitoring
TrailGuard AI cameras detected armed intruders in Similipal Tiger Reserve on 18 February 2026; the Field Director reports that at least 39 poachers surrendered and that nine guns were seized.
at least 39 poachers surrendered
Rishi Vohra wrote that Tesla's Full Self-Driving pulled his Cybertruck onto a freeway shoulder after he passed out on 18 February 2026; no vehicle data has been published.
xAI's Grok told a man in Norway, sent home with an acid blocker, to demand a computed tomography scan; his appendix, which he later wrote had ruptured, was removed on 3 December 2025.
Frankfurt police found a missing 16-year-old in crisis within a day, after the face recognition software VAS Videmo360 flagged her on a Bahnhofsviertel camera under a judge-confirmed order.
found a missing 16-year-old in crisis within a day
Daejeon's Kkumdori care robot, built on ChatGPT 4.0, flagged a woman in her 70s saying “I want to die” at 2 a.m. on 6 August 2025; police came, and a hospital admitted her the next day.
ChatGPT named an insulinoma as the likely cause of a man's three years of undiagnosed symptoms; hospital blood tests confirmed it, and surgery cured him.
surgeons removed about two thirds of his pancreas and his symptoms went
The emergency pull-over function of the NIO World Model, NIO's assisted-driving software, stopped a car whose driver fell asleep at 110 km/h on 1 June 2025, and an SOS call woke him.
Dorset Council reports that total falls fell by 49.2% and hospital transfers by 79.3% after six care homes installed Ally Labs acoustic and movement sensors.
total falls fell by 49.2%
UNEP lists eight methane leaks in Turkmenistan that were mitigated from November 2024 to March 2026 after alerts from its Methane Alert and Response System (MARS), which uses AI.
eight methane leaks in Turkmenistan that were mitigated from November 2024 to March 2026
An alert from the UN Environment Programme's Methane Alert and Response System led an Algerian operator to fix an unlit flare at Hassi Messaoud; satellites show no emissions since 15 October 2024.
no further emissions have been detected at the location since 15 October 2024
Guangming Daily reports that Wang Bing, who lost both hands, eats, drinks and writes his name with a Chongqing institute's “smart hand”, which decodes muscle signals by algorithm.
Representative Jennifer Wexton, who lost her speech to progressive supranuclear palsy, addressed the House on 25 July 2024 through an ElevenLabs clone of her own voice.
After a positive fecal occult blood test, Fujifilm AI endoscopy at St. Martin De Porres Hospital, Chiayi, marked a 1 to 1.5 cm adenoma; pathology: stage I colorectal cancer, removed endoscopically.
a 1 to 1.5 cm adenoma that biopsy confirmed as stage I colorectal cancer
The Tamil Nadu Forest Department's AI thermal cameras at Madukkarai sent over 7,116 alerts by 7 July 2026, with no elephant deaths there; a train killed two elephants nearby on 13 August 2026.
about 9,481 safe elephant crossings
Ink-detection models that three contestants built on TimeSformer recovered more than 2,000 characters of Greek text inside a sealed Herculaneum scroll, and a later scan confirmed the reading.
recovered 15 columns and more than 2,000 characters of Greek text
The NEO wireless epidural brain-computer interface let a 54-year-old man, quadriplegic for 14 years, drink unaided by driving a pneumatic glove; the research team announced this on 29 January 2024.
drink water unaided, with grasp-decoding accuracy above 90%
Neuralink's N1 implant and its decoding software let Noland Arbaugh, paralyzed after a diving accident, control a computer by thought, and he still used it 28 months after surgery.
about 10 hours of daily use for chess, coursework, scheduling and communication
Wes Ramage of Calgary, who is blind, uses Envision Glasses, a Google Glass headset with Envision's AI software, to read street signs and menus, find chairs and objects and describe scenes.
Microsoft, which funds ALERTCalifornia, says that the network's cameras and AI helped crews detect Kern County's Trotter Fire in August 2023 and hold it to 52 acres; no source names the software.
contain the Trotter Fire at 52 acres
A speech neuroprosthesis at the University of California, Davis, decoded Casey Harrell's attempted speech into text, and he used it at home to communicate 1,960,163 words.
248.3 hours of conversation over 8.4 months
Keith Thomas, paralyzed since 2020, regained arm movement and touch with the Feinstein Institutes’ double neural bypass, and its AI agent raised grasp-force success to 87% from 27%.
87% success with reinforcement learning and 27% without it
In Operación Influencer (11 May 2023) prosecutors and police detained a gang of 10 aged 15–22 behind eight accredited violent robberies; prosecutors credit the AI model Fiscal Heredia.
detained a gang of 10 aged 15–22 behind eight accredited violent robberies
In NHS Grampian's GEMINI evaluation, Mia flagged a 6 mm breast cancer that radiologists had missed in a patient named Barbara, and she had surgery and needed only five days of radiotherapy.
Mia helped find 12% more cancers than routine practice
ChatGPT listed tethered cord syndrome among three possible diagnoses after a mother entered her son's scan notes and symptoms, and a neurosurgeon found the tethered cord and released it.
Nigeria's electoral commission reported on 26 October 2022 that its Automated Biometric Identification System (ABIS) flagged 2,780,756 (22.6%) of 2021–22 registrations as ineligible.
2,780,756 registrations (22.6%) were identified as ineligible and invalidated
A deep-learning model from IBM Research and Yamagata University flagged likely sites in aerial photographs, and fieldwork in Peru confirmed 303 new Nazca geoglyphs.
303 previously unknown figurative relief-type geoglyphs, almost doubling the known total
DeepMind's deep reinforcement learning controller, commanding all 19 TCV coils, produced and held elongated, negative-triangularity, snowflake and twin-droplet plasmas at EPFL's Swiss Plasma Center.
narrowed the separation to 6.6 cm with a combined root-mean-square error of 3.7 cm
After a hit-and-run left a Singapore motorcyclist unconscious, Apple Watch Fall Detection called an ambulance and alerted his girlfriend.
Mayo Clinic's AI-ECG algorithm gave two electrocardiograms from a 41-year-old man probabilities of 97.88% and 98.67% for reduced ejection fraction, and echocardiography the next day confirmed it.
at four months his ejection fraction had improved to 55%
Gert-Jan Oskam used the STIMO-BSI interface to stand, walk, and climb stairs after a spinal cord injury.
walk more than 10 meters with crutches while the stimulation was off
Karin, a Swedish amputee, has used a bone-anchored, nerve-linked prosthesis with a Prensilia Mia Hand daily for over three years; her phantom limb pain fell from 5 to 3 and her stump pain vanished.
her phantom limb pain fell from 5 to 3
Machine-learning models using satellite imagery and phone metadata set eligibility for Togo's Novissi rural cash; 138,531 beneficiaries were paid from November 2020 to August 2021.
reduced exclusion errors by 4–21% relative to geographic alternatives
Darpan, a Telangana State Police facial-recognition tool, matched Som Soni, missing from Allahabad since 14 July 2015, to a boy in a child home in Goalpara, Assam, where his parents identified him.
Odakyu's Panasonic-built fall detection system alerts staff to people on the tracks or at the platform edge; the vendor reports that it flagged a staggering passenger on 3 April 2020.
PlantVillage's eLocust3m app gathered over 240,000 locust records, and AI removed unusable ones for the control operation that FAO credits with averting USD 1.56 billion in losses.
protected the food security of 36.6 million people and avoided USD 1.56 billion in losses
Two Tesla Model X cars stopped as an oak fell across the A31 in Storm Dennis; eight occupants walked away, and one driver credits Autopilot's emergency braking.
eight occupants walked away
On 14 October 2019 a lost hiker fell about 30 feet in Hartshorne Woods Park, New Jersey; his new Apple Watch's Fall Detection called 911 and texted his family with his location.
SkinVision rated a 31-year-old's changing arm mole high risk; stage 1A melanoma was found and removed, and he is cancer-free.
stage 1A melanoma was found and removed, and he is cancer-free
PAWS-directed ranger patrols in Srepok Wildlife Sanctuary removed over 1,000 snares in December 2018, versus 200–300 in a typical month.
removed over 1,000 snares in December 2018, versus 200–300 in a typical month
A woman with locked-in amyotrophic lateral sclerosis used the Utrecht NeuroProsthesis, an implanted brain-computer interface, for nearly seven years of independent at-home use, until March 2023.
seven years of independent at-home use
ActiveRemediation, a model built at the University of Michigan, ranked Flint homes for inspection and replacement of lead or galvanized-steel service lines, with 91.6% held-out accuracy.
held-out prediction accuracy was 91.6%
Kiribati used Global Fishing Watch to document a vessel fishing illegally in the Phoenix Islands Protected Area; a US$1 million fine plus a US$1 million goodwill grant followed.
a US$1 million fine plus a US$1 million goodwill grant followed
TIAA’s CEO said that an unnamed AI tool flagged a 76-year-old’s unusual withdrawal and that staff then kept his retirement savings of more than $3 million from moving.
retirement savings of more than $3 million
At Columbia, the EchoNext deep-learning model flagged an 84-year-old's pre-operative ECG; echocardiography then showed severe aortic stenosis, and she had a valve replacement after her surgery.
an ejection fraction of 40% and severe aortic stenosis
A scan supported by Brainomix 360 Stroke identified Jean Hines's major stroke on arrival, and she was transferred in 25 minutes for successful thrombectomy and recovered well.
transferred in 25 minutes for successful thrombectomy
Huazhong University of Science and Technology reports that 343 deaf users regained speech capabilities through practice with its unnamed AI lip-reading system.
Viz ICH Plus flagged a brain hemorrhage on a 47-year-old man's scan and alerted a transfer team, which moved him in 101 minutes against a 199.7-minute average.
101 minutes against a 199.7-minute average
Viz HCM gave a 61-year-old man's electrocardiogram a probability above 0.95 for hypertrophic cardiomyopathy; imaging confirmed the disease, and mavacamten eased his symptoms by week 8.
by week 8 the resting gradient was 9 mm Hg
SpliceAI scored a hidden PKHD1 variant at 0.98, and after RNA studies confirmed it, the parents of the infant who died used the diagnosis to plan IVF with embryo genetic testing.
a maternally inherited deep intronic PKHD1 variant scoring 0.98
The Modus V Synaptive robot’s software fused brain scans for planning and warned when the instrument strayed while a neurosurgeon removed a 4 cm clot from an awake stroke patient.
About 40 ml of clot came out in just over 30 minutes
Qure.ai's qXR flagged lung nodules on chest X-rays taken for other reasons; computed tomography and biopsy confirmed early-stage lung cancer in five adults, and four had surgery.
confirmed early-stage lung cancer in five adults, and four had surgery
In rapid genome sequencing, Fabric GEM scored a homozygous ANTXR2 variant at 4.7; clinicians diagnosed an infant within five days, and the result changed her care.
clinicians diagnosed an infant within five days
China Medical University Hospital’s Exercise ECG AI-Assisted Interpretation System gave a man a 72 percent chance of severe narrowing, and angiography found a 99 percent stenosis.
a 99 percent stenosis of the left anterior descending artery
Imagene's algorithm predicted an EGFR mutation from a stained biopsy slide; after a rapid PCR test confirmed it, the patient began osimertinib two days after biopsy.
one week earlier than routine practice and avoided whole-neuraxis radiotherapy
RapidAI's large vessel occlusion alert prompted a physician to mobilize the stroke team for Eric, 29; thrombectomy restored flow under 60 minutes from arrival; no deficits.
thrombectomy restored flow under 60 minutes from arrival; no deficits
After two years of failed phasing, a CNRS team in Palaiseau solved the structure of a protein from the mRNA decay pathway by molecular replacement with AlphaFold and RoseTTAFold models.
A Medtronic MiniMed 640G pump with predictive low glucose suspend cut hypoglycemic episodes for a woman with type 1 diabetes, who gave birth at 39 weeks after two miscarriages.
hypoglycemic episodes fell from 2.1–2.5 to 1.4 per day
Lunit INSIGHT CXR flagged nodules on chest radiographs taken for other reasons in 13 patients at Yongin Severance Hospital; all 13 had their lung cancer removed, and seven were stage I.
13 (17.3%) had cancers found incidentally because the software flagged a nodule
NICE reviewed 70 studies, most of them randomized trials, and concluded that AI polyp detection systems such as GI Genius improve the adenoma detection rate in colonoscopy.
NICE reviewed 16 studies of AI fracture software such as BoneView and found that readers' sensitivity rose without lost specificity, from 61% to 74% in one study.
from 61% to 74%
Wigan Council reports that Agilisys QuickAction tools saved an estimated 107 staff days on minutes and made 4,500 tenant survey responses usable for analysis.
NICE cites a study in which NHS services using the Limbic Access referral chatbot saw referrals rise 15% against 6% in control services, with larger rises in minority groups.
referrals rise by 15%, against 6% in control services
In FDA's review of DermaSensor, primary care physicians aided by the device referred skin cancers with 91.4% sensitivity, against 82.0% unaided; standalone sensitivity was 95.5%.
Their sensitivity rose from 82.0% to 91.4%
EMA qualified PathAI's AIM-NASH after a 1,481-case validation in which AI-assisted pathologists agreed better with expert panels on inflammation and ballooning.
hepatocellular ballooning (0.563 against 0.414)
Rekor Scout software flagged David Zayas's trips past Westchester County's 480 plate-reader cameras; officers who stopped him in Scarsdale on 10 March 2022 found crack cocaine, a pistol and cash.
In the reader study that FDA reviewed for Paige Prostate, assisted reads improved sensitivity in 7.3% of prostate biopsy images with cancer, and specificity rose 1.1 points.
improvement in sensitivity in 7.3% of biopsy images
In the randomized Italian trial that FDA reviewed, colonoscopy with GI Genius found adenomas in 55.1% of low-risk patients, against 42.0% with standard colonoscopy.
an adenoma detection rate of 55.1% with GI Genius against 42.0% with standard colonoscopy
In FDA's review of Caption Guidance, eight nurses new to echocardiography scanned 240 patients with AI guidance; cardiologists judged left ventricular size assessable in 98.8% of exams.
left ventricular size in 98.8% of exams
Overnight heart-rate alarms from an Apple Watch Series 2 prompted a woman to visit a United Kingdom hospital, where an electrocardiogram showed atrial fibrillation and she began an anticoagulant.
her pulse exceeded 120 beats per minute while she was inactive
CMS found that Viz.ai ContaCT shortens time to specialist notification for large vessel occlusion stroke; in routine use the median time to notification was 4.31 minutes.
the median time to notification was 4.31 minutes
In the reader study that FDA reviewed for OsteoDetect, 24 clinicians found distal radius fractures with 80.3% sensitivity when aided, against 74.7% unaided.
Sensitivity rose from 74.7% to 80.3%
In the reader study that FDA reviewed for QuantX, 19 radiologists reading 111 breast MRI cases raised their AUC from 0.7055 to 0.7575 with the software.
rose from 0.7055 to 0.7575
In the trial that FDA reviewed for the IDx-DR De Novo grant, the AI found more than mild diabetic retinopathy with 87.4% sensitivity and 89.5% specificity in primary care.
an observed sensitivity of 87.4%
A consumer wearable of a brand that the report does not name recorded a 59-year-old's atrial fibrillation, so that he skipped 30 days of monitoring; after ablation he was free of it without drugs.
In a Leeds Trinity University trial of 105 adults with tinnitus, 8 weeks of Woebot cut anxiety and depression scores against a waitlist; Tinnibot did more.
Woebot users lowered depression scores on the PHQ-9 from 10.03 to 8.06
In a 62-person Guangzhou trial, 8 weeks of AI pose-guided home exercise added 62.77 m to 6-minute walk distance and lowered systolic pressure more than usual care.
The app group gained 62.77 meters more in the 6-minute walk test than the control group
HM Revenue & Customs estimates that cases flagged by its Connect data-matching system yielded about £4.6 billion of additional tax in 2024–25.
£3.4 billion of additional annual yield, about £4.6 billion in 2024–25
In a UW Health randomized trial of 66 practitioners, the Abridge ambient scribe lowered work exhaustion scores by 0.44 points and cut note time by 0.36 hours a day.
Time spent on notes fell by 0.36 hours a day
ECMWF's AIFS Single forecast model went into operation on 25 February 2025; ECMWF reports 5–15% lower errors than IFS, and two independent studies in China partly confirm the gain.
5–15% lower errors than IFS
In a 368-person Johns Hopkins trial, the Sweetch AI-led diabetes prevention program met the risk-reduction goal as often as human coaching (31.7% vs 31.9%).
31.7% of the AI group and 31.9% of the human-coaching group met the combined goal
In a 2024 randomized trial with 187 dental undergraduates in China, students who used ChatGPT-3.5 with instructional videos scored 73.12 against 65.54 on a virtual reality skills test.
The ChatGPT group scored 73.12 against 65.54 for the video-only group (P < .001)
In a randomized trial with 137 law students, Vincent AI and o1-preview raised the quality of legal work and raised productivity on five of six tasks by 50% to 130%.
raised productivity on five of six tasks by 50% to 130%
In a Tanzanian trial of 2,378 people with diabetes, SELENA+ results and counselling at screening raised clinic attendance for referable retinopathy from 18.2% to 35.9%.
raised clinic attendance for referable retinopathy from 18.2% to 35.9%
The deep-learning pipeline AresGW found eight new O3 candidates with p_astro > 0.5 (cumulative 5.94); an independent burst search detected one of them, and a later search recovered none.
eight new O3 candidates with p_astro > 0.5 (cumulative 5.94)
At Islamic International Medical College, Pakistan, 100 students were randomized in 2024; after two hours of study, the ChatGPT-4 group scored 22.8 against 19.24 on an acute abdomen test.
the ChatGPT-4 group scored 22.8 against 19.24 (p = 0.005, Cohen's d = 0.58)
Rentosertib, designed with Insilico Medicine's generative-AI platform, raised mean forced vital capacity by 98.4 mL at 60 mg daily in a phase 2a trial, against a fall of 20.3 mL on placebo.
mean forced vital capacity change of +98.4 mL versus -20.3 mL on placebo over 12 weeks
In a World Bank randomized trial in Benin City, six weeks of teacher-guided Microsoft Copilot (GPT-4) tutoring raised scores 0.31 standard deviations overall and 0.23 in English.
0.31 standard deviations overall and 0.23 in English
An AlphaFold-Multimer screen predicted that Tmem81 joins a sperm complex with Izumo1 and Spaca6; the Pauli laboratory then showed that Tmem81 is essential for male fertility in zebrafish and mice.
In a UC Berkeley experiment with 274 Mechanical Turk workers in 2023, ChatGPT-written math hints gave a 17% pre-to-post gain, the only gain significant against a no-hint control.
The ChatGPT group showed a 17.00 percent learning gain from pre-test to post-test.
In a Cornwall NHS evaluation of 15 people who could not tolerate EEG, Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers; Danish validators reported 93.7% sensitivity.
Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers
A smartwatch of unstated brand notified a 25-year-old man several times of an irregular rhythm, and physicians confirmed atrial fibrillation and restored normal rhythm.
atrial fibrillation with rapid ventricular response at 123 bpm
Foncier innovant, a tool of France's tax administration that detects pools in aerial photographs, found 140,000 taxable pools in 2023, which the administration values at €40 million in local tax.
140,000 taxable pools in 2023
vivo's on-device accessibility suite offers offline speech, vision and sign language features; its sign language avatar scored 3.513 of 5 in a 2024 shared task, against 2.119 for the baseline.
In trials with 2,483 Georgia State students in online Government and Microeconomics courses from 2021 to 2023, the TA Pounce text chatbot raised the share earning an A or B by 4 points.
raised the likelihood that students earned an A or B by four percentage points
In the MASAI trial of 105,934 women in Sweden, Transpara-guided triage cut screen readings by 44.2% and raised cancer detection from 5.0 to 6.4 per 1,000.
a 44.2% reduction in workload
In the ACCESS trial, all 81 young people offered IDx-DR completed an eye exam, compared with 18 of 82 controls referred to an eye-care provider.
all 81 young people offered IDx-DR completed an eye exam, compared with 18 of 82 controls
DeepMind's AlphaTensor found ways to multiply 4x4 matrices over the field with two elements in 47 multiplications and 5x5 matrices in 96; two mathematicians then cut the 5x5 count to 95.
multiplying two 4x4 matrices over the field with two elements in 47 multiplications
At MetroHealth in Cleveland, reminder calls to patients flagged by an Epic no-show model cut no-shows from 36% to 33%, and from 42% to 36% for Black patients, in a 2022 randomized initiative.
the no-show rate was 33% with calls and 36% without
A Department for Work and Pensions machine learning model flags high-risk Universal Credit advance requests for caseworkers; the department estimates that it saved £4.4 million in three years.
DWP estimates that the model saved £4.4 million from 2022–23 to 2024–25
In a stepped-wedge trial at four Houston stroke centers, Viz.ai automated large vessel occlusion detection with phone alerts cut door-to-groin time by 11.2 minutes.
door-to-groin time fell by 11.2 minutes
At four Houston comprehensive stroke centers, Viz.ai LVO detection with mobile alerts cut door-to-groin time for thrombectomy by 11.2 minutes.
fell by 11.2 minutes (95% CI 4.2 to 18.2) after activation
In a Seoul trial of 576 patients with suspicious skin lesions, non-expert physicians using Model Dermatology diagnosed 53.9% correctly against 43.8% without it.
diagnosed 53.9% correctly against 43.8% without it
A smartwatch of unstated brand alerted a woman taking ciprofloxacin to a heart rate below 40 beats per minute; she went to hospital, and her rate recovered after the antibiotic was replaced.
by the next day her heart rate was 60–70 bpm
DGFiP's Foncier innovant deep-learning tool finds pools in aerial images; it led to more than 20,000 pools being taxed for 2022, close to €10 million for communes
more than 20,000 pools being taxed for 2022, close to €10 million of extra revenue
Researchers combined AlphaFold and RoseTTAfold structure predictions with electron microscopy maps to model more than 90% of the human nuclear pore scaffold, about twice the previous coverage.
covered more than 90% of the scaffold, roughly twice the coverage of the previous model
In the RAIDERS trial at four Kigali diabetes clinics, immediate Cybersight AI feedback raised 30-day referral attendance to 51.5% from 39.6% with delayed human grading.
raised 30-day referral attendance to 51.5% from 39.6% with delayed human grading
Nigeria used machine-learning poverty maps from satellite imagery to select 1,163 of its poorest urban wards for a COVID-19 cash transfer that counted 49,000 beneficiary households by 15 October 2021.
49,000 beneficiary households as of 15 October 2021
A GPT-based chat assistant raised issues resolved per hour by 15% on average among 5,172 support agents, with the largest gains for novice and low-skilled agents.
raised issues resolved per hour by 15% on average
A Samsung Galaxy Fit band recorded a pulse of 188 beats per minute in a woman in South Korea, and with a routine electrocardiogram the record supported her referral for a successful ablation in 2020.
recorded a peak resting pulse rate of 188 beats per minute
In the PROSPECT trial at a Blantyre clinic, TB patients offered HIV testing plus chest X-rays read by CAD4TB started treatment a median of 1 day after the visit, against 11 days under standard care.
Median time to treatment was 1 day in the X-ray arm and 11 days under standard care
In a 2019 trial in 178 public high schools in Espírito Santo, Brazil, Letrus automated essay scoring and feedback raised ENEM essay scores by 0.094 standard deviations.
raised essay scores by 0.094 standard deviations
Upstart's machine-learning lending model approved 15 to 30 percent of low-score applicants that traditional models reject, at lower rates; a 2024 monitor found disparities for Black applicants.
approves 15 to 30 percent of low-score applicants rejected by traditional models, at lower rates
In the HYPE trial at Amsterdam UMC, the Hypotension Prediction Index cut median time in hypotension during surgery from 32.7 to 8.0 minutes per patient.
median time in hypotension fell from 32.7 to 8.0 minutes per patient
In an 80-school trial in Rajasthan, India, students in the 40 schools that used Mindspark in school hours scored 0.22 SD higher in math and 0.20 SD higher in Hindi after 18 months.
0.22 standard deviations higher in math and 0.20 standard deviations higher in Hindi
Researchers estimated an 82% (75%–89%) relative cut in eagle deaths at a Wyoming wind farm after IdentiFlight began curtailing turbines; two federal statisticians later put it at 50% (−159%, 89%).
82% (75%–89%) relative cut in eagle deaths at a Wyoming wind farm
In a 2018 Memorial Sloan Kettering randomized trial of 683 surgeries, a machine learning model cut scheduling error by 9.8 minutes and patient waits by 33.1 minutes.
fell by 33.1 minutes per patient, from 49.4 to 16.3 minutes
Dexcom's analyses of 1,424 and 15,000 users link the G6's 20-minute Urgent Low Soon alert to less hypoglycemia; authors of an independent trial in 24 adults found a reduction only below 2.8 mmol/L.
time below 54 mg/dL fell from 1.0% to 0.6%
In a randomized crossover home trial, eight people with above-elbow amputation and nerve transfer surgery did better on two hand-function tests with pattern-recognition control; seven preferred it.
moved three clothespins in 90.2 seconds on average against 137 seconds with direct control
AdmitHub’s text-message assistant Pounce raised enrollment by 3.3 percentage points among students already committed to Georgia State in 2016.
raised enrollment by 3.3 percentage points among students already committed to Georgia State
In a 71-child trial, Stanford's Superpower Glass emotion classifier on Google Glass added 4.58 points on a socialization scale, and Spectrum reported that the gain was gone six weeks after treatment.
gained 4.58 points more on the Vineland socialization subscale than controls (P = .005)
Students offered free access to Mindspark centers gained 0.37 standard deviations in math and 0.23 in Hindi over 4.5 months.
0.37 standard deviations in math and 0.23 in Hindi over 4.5 months
In independent studies of 6 to 100 participants, people scored higher on reading and daily tasks with OrCam readers; in a Dutch study only three of 20 patients kept using the device.
improved on a 10-item daily-living test from 2.5 to 9.5
TIGTA found that IRS Return Review Program identity-theft models alone selected $1.88 billion, or 47.8%, of $3.92 billion in confirmed identity-theft refunds in 2015.
selected $1.88 billion, or 47.8%, of $3.92 billion in confirmed identity-theft refunds
eBay's in-house machine translation, introduced for Latin America in May 2014, raised exports on eBay by 10.9% in the published study and 17.5% to 20.9% in the working paper.
raised exports on eBay by 10.9% in the published study
Students using Cognitive Tutor Algebra I scored 0.21 standard deviations higher in the second year, with no statistically significant difference after correction for multiple comparisons.
scored 0.21 standard deviations higher in the second year
After Waymo detected a firearm violation, one of its driverless cars pulled over in San Francisco on 3 September 2026; police found a loaded ghost gun and booked two minors.
On 13 June 2026 Phú Thọ police saw teenagers preparing weapons to fight in Việt Trì, and the provincial AI camera system identified all 18 faces in about 30 minutes.
identified all 18 faces in about 30 minutes
On 2 May 2026 ZeroEyes gun detection flagged a man pointing a gun in a Florida State University garage, and police found a stolen Glock 19 on him and charged him, court documents say.
In an 18-week trial, Origin Orbit flagged suspected leaks from satellite data, and Thames Water found over 800 leaks, saving an estimated 8.7 million liters a day by the utility's figures.
over 800 leaks across the network, saving an estimated 8.7 million liters of water a day
Amazon Mining Watch uses an AI detector to map gold mining, and its report shows 496,000 hectares affected between 2018 and 2025 in protected areas and Indigenous territories of the Amazon.
496,000 hectares affected between 2018 and 2025
The Methane Alert and Response System of the United Nations Environment Programme detected a leak in Alberta on 10 April 2026, and the operator repaired it on 17 April 2026.
released 175 tonnes of methane in a single week
President Zelenskyy said on 13 April 2026 that Ukrainian ground robots and drones had taken an enemy position with no infantry and no Ukrainian losses; no source has verified the claim.
more than 22,000 missions on Ukraine's front line
In 2025, its first season with AI models in operations, the US National Hurricane Center found that Google DeepMind's ensemble mean beat its official Atlantic track forecasts at 12 to 72 hours.
beat its official Atlantic track forecasts at 12 to 72 hours
An AI smoke-detection camera of unnamed make, installed for Arizona Public Service, spotted early signs of the Diamond Fire near Flagstaff on 26 March 2026, and crews held the fire to 7 acres.
contained it before it grew past 7 acres (2.8 hectares)
A BrainGate2 implant with a recurrent neural network decoder turned two paralyzed men's attempted finger movements into typing, at up to 22 words per minute with a 1.6% word error rate.
T18 reached 110 characters (22 words) per minute with a 1.6% word error rate
Wiz researchers used Claude Code and IDA MCP to reverse engineer GitHub's binaries and found a critical remote code execution flaw, which GitHub fixed on github.com the same day.
went from idea to working exploit in under 48 hours
After alerts from UNEP's Methane Alert and Response System, which uses AI, operators diverted gas from an Iraqi source by 25 February 2026; UNEP estimates that it emitted 60,500 tonnes of methane.
an estimated 60,500 tonnes of methane emissions
Istanbul police first used their AI-supported AVCI messaging-analysis system on 10 February 2026, in a drug operation across 14 provinces: 325 detained and remanded, 69 groups closed.
325 detained and remanded, 69 groups closed
The AI door god camera system from LEADERG flagged smoke in an empty office at Beigang Wude Temple before a battery fire, and the fire burned ten square meters with nobody hurt.
the fire burned ten square meters with nobody hurt
After withdrawals were refused, a 77-year-old Bengaluru man checked the firm with ChatGPT, which flagged complaints and inconsistencies, and he reported a Rs 67.10 lakh fraud to police.
Ghana's Finance Minister reported that the Publican AI Trade Solution raised assessed customs collections by over US$300 million from January to 17 July 2026, a 17.5 percent uplift.
raised assessed customs collections by over US$300 million from January to 17 July 2026
LNER reports that One Big Circle's Automated Intelligent Video Review (AIVR) system detected a minor track fault near Retford, which engineers repaired overnight with no delay to passengers.
an overnight repair with no delay to passengers and zero delay minutes
The Ho Chi Minh City police's unnamed street AI camera system identified a masked burglar who took two laptops and 14 million VND on 23 December 2025; police detained him on 24 December.
Garmin Autoland engaged when a King Air B200 lost cabin pressure on 20 December 2025 and landed it at Rocky Mountain Metropolitan Airport, with the two pilots unhurt.
two pilots unhurt
Paul Conyngham used ChatGPT and AlphaFold to choose targets for an mRNA vaccine; with a checkpoint inhibitor, it halved Rosie's largest tumor by early 2026 before the cancer returned.
halved Rosie's largest tumor by early 2026
Open Climate Fix's machine-learning solar forecast, Quartz Solar, is live in the control room of Great Britain's grid operator; the maker claims at least £30m a year in avoided imbalance costs.
at least £30m a year in avoided imbalance costs
The US National Hurricane Center blended Google DeepMind's AI ensemble into its Melissa forecasts from 21 October 2025; the center's track errors at 120 hours were less than half its five-year mean.
track errors at 120 hours were less than half its five-year mean
The operators said that Veesion's gesture-recognition software flagged a shopper hiding goods in their Szeged store at night, and that police caught the perpetrator within hours.
over four and a half months the system had exposed three theft attempts
Bug reports made mostly with the ZeroPath AI code analyzer led curl to merge about 50 fixes from a first list, and by 23 December 2025 AI analyzer reports had led to several hundred.
fixed several hundred bugs as a direct result of reports from AI analyzers
A former Cardiff teacher entered her symptoms into ChatGPT after a coma; it listed hereditary spastic paraplegia, which genetic testing confirmed in August 2025 (BBC, WalesOnline).
Image-analysis software, which its developer TEXSAR identifies as ADIAT, flagged a red helmet in about 2,600 drone frames, and rescuers recovered Nicola Ivaldo's body on Monviso by 31 July 2025.
located at about 3,150 m and recovered by 31 July 2025
Giovanni Abitante used unnamed generative video tools to animate archival photographs for I Am an American, a short film that received 2025 awards; no source measures its effect on viewers.
ChatGPT listed a blood pressure drug among possible causes of a woman's 18-month cough; a doctor confirmed it and changed the drug, and her daughter says that she is healing.
A Reddit user says that ChatGPT tied ten-plus years of symptoms to an MTHFR result and proposed B12, and that the symptoms largely resolved; a doctor later doubted that the case was real.
The Taichung Environmental Protection Bureau's AI smoke-recognition system flagged roadside waste burning in Dadu District, and police identified a 74-year-old repeat burner, who faces a fine.
a fine of up to NT$100,000
Google says that its Big Sleep agent, given signs that attackers were preparing to exploit an unknown flaw, isolated SQLite flaw CVE-2025-6965, which SQLite fixed in version 3.50.2 on 28 June 2025.
A prototype of Sky Sentinel, a Ukrainian AI-aimed machine-gun turret, downed six Shaheds in six uses, a unit commander said; more than 12 turrets were on duty in Kyiv by October 2026.
downed six Shaheds in six uses
OpenAI's o3, which the researcher Sean Heelan ran through the plain application programming interface, flagged a previously unknown use-after-free in the Linux kernel's SMB server, now CVE-2025-37899.
The Karnataka Examinations Authority's facial recognition app flagged a woman with a forged entrance-test admit card in Bengaluru on 17 April 2025; she fled before writing the paper.
Israeli facial recognition software, unnamed in the reports, matched a robbery suspect's camera image to a police record; two arrested within about a day and Rs 79.5 lakh of the Rs 80 lakh recovered.
two arrested within about a day and Rs 79.5 lakh of the Rs 80 lakh recovered
Şırnak police stopped a car selected by the ASENA narcotics analysis software; the dog Hector found 86.6 kg of skunk in 119 packages in the trunk, and three lawyers were remanded (March 2025).
86.6 kg of skunk in 119 packages
A police facial recognition system, which no report names, identified a man stopped in Adana without identification as a robbery convict wanted to serve a 10-year sentence.
a fugitive convict wanted to serve 10 years, 1 month and 10 days for robbery
A Tokat school bus driver used ChatGPT to write his objection to a 2,400 TL fine for dropping a student off outside a stop; the court annulled it within three weeks for lack of evidence.
The CGU's ALICE tool flagged three 2024 federal hospital tenders in Rio de Janeiro state; auditors confirmed R$19.4m of overpricing risk, and one was suspended, one revoked and one re-estimated
a saving of R$3,063,774.84
The Metropolitan Police's Live Facial Recognition cameras identified David Cheneler, a sex offender, alone with a six-year-old girl on 10 January 2025, and he was jailed for two years.
Kingston Crown Court jailed him for two years
Skylight reports that its speed alert and its machine-learning fishing alert led Argentina's coast guard to intercept a trawler on 5 January 2025, and that the vessel paid a fine.
A mother posted on 11 January 2025 that Grok on X had flagged a displaced wrist fracture that urgent care missed; an orthopedist confirmed the fracture and cast it.
OpenAI's o3 Deep Research helped Boston Children's Hospital reanalyze 376 locally unresolved rare-disease cases, which yielded 18 confirmed diagnoses: 11 new and seven rediscoveries.
18 confirmed local diagnoses, or 4.8%
Orca AI and NorthStandard report 52% fewer severe close encounters across 139 vessels of several types, comparing early and late use without an untreated fleet.
a 52% reduction in high-severity close encounters
Baidu reports that Apollo Go robotaxis give visually impaired passengers audible arrival alerts, automatic unlocking, and voice controls, and that these passengers took more than 200 rides in Wuhan.
Code Intelligence's Spark, an AI agent that writes and runs fuzz tests, found a heap use-after-free in wolfSSL in late October 2024; wolfSSL shipped the fix in 5.7.6 on 31 December 2024.
The Scott-Morgan Foundation reports that its AI communication platform enabled Erin Taylor, who has ALS, to introduce Lenovo's CEO at Tech World on 15 October 2024.
enabled Erin Taylor, who has ALS, to introduce Lenovo's CEO at Tech World
The U.S. Treasury reported that machine learning helped it recover $1 billion in check fraud in fiscal year 2024, part of over $4 billion prevented or recovered; no source names the software.
$1 billion was recovered through machine-learning identification of Treasury check fraud
GiveDirectly used Google Flood Hub forecasts to trigger $105 payments to over 4,600 people in Ibaji, Kogi State, most of them before floods peaked on 21 October 2024.
weekly incomes rising from ₦4,500 to ₦10,000 and food shortages falling from 97% to 7%
SUVAS, the Supreme Court of India's translation software, translated 36,271 Supreme Court judgments into Hindi and 17,142 into 16 languages by 5 August 2024
translated 36,271 Supreme Court judgments into Hindi and 17,142 into 16 languages
A Pano AI camera funded by Xcel Energy alerted crews to a campsite fire near a Breckenridge neighborhood; firefighters said that the early notice kept it to one-eighth of an acre.
the early notice kept it to one-eighth of an acre
DigitalPath's smoke-detection software on ALERTCalifornia cameras flagged a fire near Auburn at 2:33 a.m. on 6 July 2024, before any 911 call, and crews kept it to a 30 x 30 foot area.
Crews put the fire out before it spread beyond a 30 x 30 foot area
Utility-funded Pano AI cameras detected smoke from the 16 June 2024 Bear Creek Fire in Douglas County, Colorado; helicopters were dispatched within 30 minutes and the fire was held to three acres.
helicopters were dispatched within 30 minutes and the fire was held to three acres
Tutor CoPilot, a Stanford system built on GPT-4 that suggests moves to human tutors, made students 4 percentage points more likely to master topics, with no significant gain on end-of-year tests.
4 percentage points more likely to master topics
Cheshire Constabulary's facial recognition system matched camera footage of an unknown Northwich dealer to Keenan Hughes; he and Vincent Potter were stopped on 17 May 2024 and jailed on 1 July 2024.
jailed Hughes (18) for 28 months and Potter (19) for 34 months
Te Hiku Media's te reo Māori speech recognition, trained on 310 hours of community recordings, went from 86% accuracy in its first model to a reported 92% in 2024.
transcribes te reo Māori with 92% accuracy
SES Water's Aquasuite Burst compares data from 1,000 flow and pressure sensors with predicted levels and raises leak alarms within minutes; the utility reports that run times fell by up to 40%.
reduced leakage run rate times by up to 40%
The LAARMA machine-learning system detected cassowaries with 97% recall and 77% precision, and drivers slowed by an average of 6.30 km/h and 5.06 km/h at two trial zones when its signs warned them.
drivers slowed by an average of 6.30 km/h and 5.06 km/h at two trial zones
DeepMind's FunSearch, which pairs a large language model with an automated evaluator, produced a cap set of size 512 in dimension 8 and raised the asymptotic lower bound for cap sets to 2.2202.
cap set of size 512 in dimension 8, exceeding the previous best known construction of 496
On 30 November 2023 ZeroEyes gun detection alerted Hobbs police to a gun seen by a camera in the Del Norte Pool parking lot, and officers detained people shooting clays, its maker reports.
charges were filed against one
A Los Angeles police drone observed a sleeping man with a suspected gun on a bus; officers then sent the Boston Dynamics Spot robot in as a barrier, and the gun proved to be a BB gun.
An AI camera program built by Shenzhen University students with Tencent engineers alerts a Shantou goose cooperative to sick geese, and Xinhua reports that survival rose 30 percent.
helped the farm increase survival of lion-head geese by 30 percent
A U.S. Customs and Border Protection machine-learning model flagged a car's crossing history at San Ysidro in 2023, and officers found over 75 kilograms of narcotics and arrested two people.
found over 75 kilograms of narcotics
ALERTCalifornia's AI flagged smoke on a Cleveland National Forest camera at about 3 am on 5 August 2023, and Cal Fire held the verified fire under a quarter acre.
Cal Fire held the verified fire under a quarter acre
The HelioLinc3D algorithm linked sparse detections from the ATLAS survey into the potentially hazardous asteroid 2022 SF289, which the Minor Planet Center published on 18 July 2023.
about 600 feet across with an Earth MOID of 0.0015 AU
ChatGPT Plus running GPT-4 listed immune-mediated hemolytic anemia among causes of a dog's worsening anemia; a second veterinarian confirmed it, and the owner reported an almost full recovery.
After the 2023 Türkiye earthquakes, xView2, a machine-learning model that grades building damage in satellite images, gave responders maps that UN rescue teams used in Adıyaman.
would typically take weeks and now takes hours or minutes
An alert from Instagram's self-harm detection led Madhya Pradesh police to a 17-year-old in Satna within about an hour on 3 November 2022, and she was counseled and sent home.
A Turkish Brain Project algorithm on Gazi University’s hospital network alerted physicians to a mass in a woman’s brain during her scan, and a biopsy the next day confirmed a glioma.
The North Rhine-Westphalia Interior Ministry said that Palantir-based DAR traced an emailed amok threat in Siegen in fractions of a second and identified 51 of about 100 people at a Duisburg shooting.
51 of the roughly 100 participants
A speech neuroprosthesis built at the University of California, San Francisco, decoded Ann Johnson's attempted speech into text at a median 78 words per minute and drove a voice and an avatar.
decoded text at a median 78 words per minute
Women at Kenyan facilities that introduced PROMPTS, a text-message platform whose helpdesk sorts questions automatically, had improvements of 0.06–0.09 standard deviations across five measures.
improvements of 0.06–0.09 standard deviations across five measures
A Stanford speech neuroprosthesis used a recurrent neural network and a 125,000-word language model to decode Pat Bennett's attempted speech into text at 62 words per minute.
produced text at 62 words per minute
The From Numbers to Names face search found Geddy Lee's mother in a photograph from the Bergen-Belsen displaced-persons camp, and survivor Blanche Fixler recognized herself in photos that it found.
raised the number of identified people in the photo from three to six
Kenneth Van Hoesen says that his Tesla Model 3 straightened its course after he swerved around a wild boar in Belgium on 30 May 2020; press reports credit Tesla's Autopilot.
Google's deep-learning system, ARDA, graded eye photographs in real time at nine Thai primary-care sites with 94.7% accuracy for vision-threatening diabetic retinopathy; 2,412 patients were referred.
For vision-threatening retinopathy the system's accuracy was 94.7%
Bret Peters's Model S on Autopilot braked and moved right as a wrong-way car came at him on Highway 49 in North Carolina, and TeslaCam recorded the near miss.
A speech neuroprosthesis built at the University of California, San Francisco, used deep-learning models to decode a paralyzed man's attempted speech at a median 15.2 words per minute.
sentences at a median 15.2 words per minute with a median word error rate of 25.6%
A speech recognizer built at Reykjavik University has drafted transcripts of speeches in Althingi, the Icelandic parliament, since January 2019, and transcribers graded 77% of transcripts as Good.
transcribers graded 77% of transcripts as Good
PlantVillage says that its Nuru app helped a Kenyan farmer select healthy cassava cuttings and that her harvest rose to 2.0–5.0 kg per plant against 0.5 kg; no independent source confirms it.
2.0–5.0 kg per plant against 0.5 kg
CBP's new facial comparison system at Dulles reported a mismatch for a man presenting a French passport on 22 August 2018; a Congolese ID was found in his shoe and he was returned to Brazil.
by September 2021, more than 900 impostors nationally
A deep convolutional neural network built by Google and the University of Texas at Austin found Kepler-90i, an eighth planet around the star Kepler-90, in archival Kepler data.
a rocky planet about 30 percent larger than Earth with a 14.4-day orbit
Clinatec's adaptive decoder turned brain signals from a man with tetraplegia into avatar and exoskeleton movements, with 64.0% success at home and 70.9% in the laboratory.
64.0% success using the avatar at home and 70.9% with the exoskeleton in the laboratory
The Spanish National Police's VeriPol text classifier flagged 25 false robbery reports in Murcia and 39 in Málaga in a June 2017 pilot week; the police stopped using it on 21 October 2024.
25 and 39 false robberies respectively, against 3 and 12 in the same month
Alice, the federal audit court's tender-reading robot, emailed tender information to auditors, who found road-maintenance notices that overlapped a concession; the court ordered them annulled.
ordered cancellation of the BR-153 tender budgeted at R$20 million
Bank F's network analytics tool, which the report does not name, linked a third company to a suspected money laundering network through shared phone numbers and email addresses.
linked a third company to a suspected money laundering network
NCAR's adaptive forecast-blending system (DICast), used by Xcel Energy since 2009, cut mean wind forecast error from about 18% to 10.8% by 2014; Xcel estimated US$49m saved (IEEE Spectrum 2015).
cut mean wind forecast error from about 18% to 10.8% by 2014; Xcel estimated US$49m saved
West Yorkshire Police reported three arrests and six Sexual Harm Prevention Order checks after NEC NeoFace live facial recognition issued nine alerts in Huddersfield on 12 to 14 August 2026.
three arrests and six Sexual Harm Prevention Order checks
Harvard and UCSF researchers co-folded 10,000 nanobodies with MRGPRX2 in AlphaFold-Multimer; of ten tested, three bound at 20 to 200 nM and blocked mast cell degranulation.
bound MRGPRX2 with affinities between about 20 and 200 nM
Southern Water's machine-learning sewer monitors flagged a forming fatberg in Rowner, Gosport, and a crew cleared it before any overflow into the River Alver.
Bradford Council reports that its Annie digital assistant handles about 50 adult-social-care conversations weekly, 20% of them outside office hours, with 74% user satisfaction.
Leeds City Council reports that planning officers using the Xylo Core AI workspace processed over 40% more householder and minor applications each month.
Zhejiang Lab and the National Astronomical Observatories of China report that OneAstronomy, built on Qwen2.5-72B, scored higher than comparison models on two astronomy test sets.
South Wales Police's Operator Initiated Facial Recognition phone app matched a man in Cardiff to a European arrest warrant from Poland dating from 2012, and he was remanded in custody (July 2025).
Mato Grosso's security secretariat reports that Vigia Mais MT facial recognition matched a woman wanted for drug trafficking in central Cuiabá on 12 July 2025, and police arrested her.
Face2Gene's DeepGestalt facial analysis changed geneticists' testing decisions in 17 retrospectively selected cases involving 19 patients, leading to confirmed diagnoses.
17 cases involving 19 patients
MEDIAIPLUS reports that its MediC and FiCRO tools cut trial-data alignment time by 85% and outsourcing costs by 10% and raised trial success by 25% in early deployments.
In the makers' randomized trial, students taught with a digital patient built on the ChatZOC retrieval framework scored 10.5 points higher in history taking; clinical benefits remain intended.
Huawei and Yunnan Baiyao present the Leigong model for traditional Chinese medicine and report faster herbal sorting and lower Poria return costs; health benefits remain proposed.
Alibaba DAMO Academy reports that its Baguan weather model had 57.8% lower regional forecast error on average than ECMWF-IFS, the European Centre for Medium-Range Weather Forecasts' system.
The team behind the Polar Eco-Carbon Intelligence Platform reports that AI-guided microalgae cultivation raised carbon dioxide absorption by 30%, against an unstated baseline.
a 30% improvement in carbon absorption
China Telecom reports that its Generative Artificial Intelligence Content Security Detection and Protection Platform screens generated material with error rates of about 6%.
false-positive and false-negative rates of approximately 6%
China Telecom Cloud reports that CloudTop, built on China Telecom's TeleChat2 and Huize models, raised the coverage of quality inspection from 10% to 100% and cut night shifts by 30%.
coverage of quality inspection from 10% to 100%
China Telecom reports that coordinated large and small AI models raised configuration compliance above 95% and cut anomaly detection from ten minutes to one.
China Telecom reports that its Hui Xin assistant cut key process errors by over 50% and raised marketing success by over 30% in two service halls.
The ITU entry reports that operations aided by a telecom foundation model, a transformer with 38 billion parameters, lowered packet loss and delay at MTN without isolating the model's share.
lowered packet loss and delay at MTN
China Telecom reports that its AI-powered control and operations platform saves more than 900 million kWh a year and has cut CO₂ emissions by more than 500,000 tons.
saves more than 900 million kWh a year
PixVerse reports that a TikTok creator in Brazil used its AI video generator to gain 300,000 followers and over 210 million views in two months.
gain 300,000 followers and over 210 million views in two months
Crowdworks AI reports that Safety Talk, a translation app for construction and shipbuilding sites built on Gemma-2-9B-IT, reached over 90% accuracy and user satisfaction in field tests.
China Southern Power Grid's AI company reports that its Intelligent Transmission Section Flow Control Agent outperformed human dispatchers in control evaluations on a regional grid.
ZTE reports that its Nebula vision system at Taicang Port eliminated 92% of false fire alarms and detected unsafe behavior; no source reports injuries prevented.
SmartMore reports that, in one classification test, the training that IndustryGPT chose reached 100% accuracy in 5:27, compared with 50% in 9:31 for manual training.
ZEZEDU reports that students improved after one month of using its CherryPot math assistant and that internal benchmark scores rose about 17 percentage points, with no independent evaluation.
DP World reports that CargoWaves, which recommends haulers with an XGBoost classifier, raised cargo throughput by 30% and lowered carbon dioxide per shipment by 20%.
At Mayo Clinic, 11 bimodal cochlear implant users scored 79% on sentences in babble with the Phonak Sphere Infinio DNN program, against 60% with a program without noise reduction.
bimodal sentence scores on the AzBio test reached 79% with the DNN program against 60%
Google's Android Earthquake Alerts detected the Marmara Ereğlisi earthquake of 23 April 2025 and alerted over 16 million users, but its stronger alert missed inland areas of moderate shaking.
Alerts reached more than 16 million Android users
A Shandong University team used scaffold domains designed with RFdiffusion and ProteinMPNN to solve three small-protein cryo-EM structures at 2.29 to 2.50 angstroms, and a fourth without them.
three small-protein cryo-EM structures at 2.29 to 2.50 angstroms
Lancashire County Council reports that its Future Mobility Platform, which combines Aimsun, VivaCity and Alchera tools, has informed bus-route planning; no measured outcome is reported.
Tunbridge Wells Borough Council reports that AI masking and editing in Adobe Lightroom made The Amelia Scott museum's scan and image processing 50% faster.
The Rising Hope Girls' Educational Foundation reports that girls in Abuja schools used generative tools such as ChatGPT and DALL-E to make art and advocacy projects.
In a 2025 provincial pilot, a system that weighs pigs by AI without contact cut shipping age on six Gyeonggi farms from 182.3 to 176.7 days and raised income per head 18%.
Income per head rose from 100,000 to 118,000 won, up 18%
KT Commerce says that Nego-Wiz, AI O2O's price negotiation software, automated 60,000 tasks and saved 8,300 hours in partial use during 2024.
St Helens Borough Council built tools in Microsoft Copilot Studio that match care assessments to a catalog of care technology and draft care plans; resident benefits remain intended.
Kunshan police say that their “AI police” models found a mobile gambling ring, broken up on 10 September 2024, and traced a telecom fraud so that 500,000 yuan was frozen within 10 minutes.
traced three tiers of card transfers and froze 500,000 yuan within 10 minutes
ALICE, the federal comptroller's AI alert tool, flagged a highway tender in Pará, Brazil, and the auditors' findings led the contracting agency to cut almost R$100 million from the tender's budget.
cut the reference value from R$549.66 million to R$436.47 million
Uppsala and Karolinska researchers docked compounds against AlphaFold models of TAAR1; 18 of 30 tested were agonists (60%), against 22% from a homology model.
Eighteen AlphaFold compounds were TAAR1 agonists, a hit rate of 60%
In a field experiment with 776 P&G professionals, individuals using a GPT-4-based tool raised solution quality by 0.37 SD and spent 16.4% less time, matching teams without AI.
raised solution quality by 0.37 SD and spent 16.4% less time
On Kangaroo Island, eVorta AI found 97.9% of 492 feral cat images against 95.3% for staff; with 4G cameras, the board estimates A$2.15 million saved over three years.
eVorta correctly classified 482 of 492 cat images (97.9%) against 469 (95.3%) for staff
UCSF-led teams docked up to 1.6 billion molecules against AlphaFold2 models of the sigma-2 and 5-HT2A receptors; hit rates (54% and 26%) matched those from experimental structures.
a hit rate of 54%, against 51% for the crystal structure
Hertfordshire County Council reports that the Robotiz3d ARRES Prevent robot, which uses AI to detect road defects, filled a small number of cracks in a public-road pilot.
Somerset Council reports that Microsoft Copilot saved pilot users two to four hours a week and that 75% reported better focus and 53% found work more enjoyable.
KoBold Metals used the Intelligent Prospector drilling planner to help appraise the high-grade zone of the Mingomba copper deposit in about eight months; no source isolates the effect of AI.
appraised the high-grade zone in about eight months
A 72-year-old man with a paralyzed replanted arm regained active thumb and wrist movement after 18 sessions in which the Neuromotus system decoded his intended movements.
active thumb extension (4 degrees) and active wrist flexion (6 degrees)
HSE University reports that a language-model pipeline turned eight million Russian-language articles into a database of 38,000 natural-hazard events.
China Mobile reports that its AI and 5G-A monitoring system widened the monitored area of a Xiamen reserve from 120 to 330 square kilometers and cut collision-related animal deaths by 97%.
widened the monitored area of a Xiamen reserve from 120 to 330 square kilometers
Researchers from Next Step Fusion and General Atomics report that reinforcement-learning controllers trained in simulation kept DIII-D plasma shape errors within 1.5 to 2.0 cm.
The Mohammed Bin Rashid Housing Establishment reports that its Smart Housing Experience, which uses AI to screen applicants, cut application approvals from 12 months to 15 days.
SK Telecom reports that its CareVia vision AI cut behavior-recording time from three hours to five minutes; its reference for 76.5% fewer self-harm incidents names no such study.
Transport for West Midlands reports that VivaCity sensors detected 61,000 near misses and that near misses at one Coventry junction fell by more than 88 percent after a refuge was built.
ChatGPT suggested the ALMS1 gene for a 44-year-old man with hearing loss and blindness since infancy and heart and kidney disease; a geneticist confirmed Alström syndrome in two business days.
confirmed Alström syndrome in two business days
Seoreu reports that WIMPLE, its AI-powered recycling bin, made recycling collection possible in Busan public spaces and improved collection scheduling.
Buckinghamshire Council reports that Microsoft Copilot cut call wrap-up by an average of two minutes a call and helped a deaf colleague take part in meetings more independently.
An electrocardiogram app on a smartwatch of unstated brand helped a telemedicine physician identify atrial fibrillation in a seafarer and recommend a ship’s diversion to the nearest port.
A click detector built on a long short-term memory network let a man with ALS spell through a brain implant at a median 10.2 characters per minute for 90 days without retraining, in the laboratory.
median spelling rate of 10.2 characters per minute
In a preregistered experiment with 758 BCG consultants, GPT-4 users completed 12.2% more tasks 25.1% faster with higher quality, but erred more on a task beyond its reach.
completed 12.2% more tasks, finished 25.1% more quickly
VitorIA identified at least 247 similar appeals in the controversy behind Tema 1.261; STF Plenary found no general repercussion on 22 August 2023 (final 23 September 2023)
identified at least 247 similar appeals in the controversy behind Tema 1.261
In a preprint coauthored by Rising Academies staff, Ghanaian students who used its WhatsApp tutor Rori one hour a week in 2023 gained 0.36 standard deviations more in math than controls.
an effect size of 0.36 (Cohen's d, p < 0.001)
A deep reinforcement-learning controller from Princeton kept DIII-D fusion plasmas below a tearing-instability threshold; all but one shot disrupted before it took control, and only two did after.
all but one shot disrupted before it took control, and only two did after
Dubai Culture and Khalifa University report that machine learning on remote-sensing data detected known excavation areas at Saruq Al-Hadid and predicted new zones that await excavation.
Camera-based computer vision from an unnamed maker detected sidewalk riding on Spin e-scooters in Santa Monica; its feedback cut time ridden on sidewalks by 22% in a randomized test.
The feedback group spent 22% less time and 20% less distance on sidewalks
Harvard students who used PS2 Pal, powered by GPT-4-0613, learned more in less time than students in an active-learning class.
median learning gains over double those of the class session
Morphing I reports that its Morphing-BOT robot inspected more than 30 km of water mains in four South Korean municipal pilots and produced validated inspection data.
In a randomized crossover trial in Brussels, newly implanted patients fitted with the FOX software understood speech better at 3 and 12 months, and all manual-arm patients kept the FOX map.
100% of the manual group kept the FOX map and 82% performed better with it
The Minas Gerais court of accounts screened 2,297 tenders from August 2022 with Alice, a system that analyzes tender notices, and obtained corrections of notices worth R$91,723,011.31.
the audit achieved corrections of notices worth R$91,723,011.31
A machine-learning damage forecast from 510 met the early-action trigger on 27 October 2022, and the Philippine Red Cross installed shelter kits for 78 Aurora households before Storm Nalgae.
installed shelter kits for 78 Aurora households before Storm Nalgae
Cerebral's CMD-1 crisis-message detector screened 102,471 patient chat messages in October 2022 and cut the daily median time to crisis-specialist triage to 8 to 13 minutes from over 9 hours.
the daily median time from message receipt to crisis-specialist triage was 8 to 13 minutes
In a Bihar trial, volunteers relayed Google and Central Water Commission flood alerts, and treated households under severe flooding were 12.7 percentage points less likely to have their home flooded.
12.7 percentage points less likely to have their own home flooded
Baidu reports that its XiLing avatar gave round-the-clock sign-language coverage of the 2022 Winter Olympics; a Deaf professor's team found that Olympic avatars lost key information.
Face2Gene listed a DYRK1A-related intellectual developmental disorder among four candidates for a man diagnosed with Cornelia de Lange syndrome at age four, and exome sequencing confirmed it.
In 2022, CAD4TB version 7 scored ultra-portable chest X-rays of 94,694 people in eight Nigerian states, alongside symptom screening: 2,337 TB cases diagnosed, 97% treated.
2,337 TB cases diagnosed, 97% treated
AIZEN reports that CreditConnect-EV has helped partner lenders finance more than 20,000 electric two-wheelers in Indonesia since 2022.
Two transulnar amputees given Coapt pattern-recognition prostheses on osseointegration implants regained activities of daily living.
With its ALICE tool, the CGU cut R$2.38m from a 2021 Ipanema hospital glove tender and R$5.95m from a glove tender at Brazil's National Cancer Institute; R$8.33m was avoided
prevented at least R$8.33 million of overpriced purchases
mMitra’s call scheduler reduced cumulative engagement drops by 32% compared with regular messaging alone over seven weeks.
reduced cumulative engagement drops by 32.0% compared with messaging alone
Two men with tetraplegia used a wireless BrainGate brain-computer interface at home; its decoders turned neural activity into cursor movement and clicks for browsing and typing on a tablet.
T5 typed 196 correct characters in 14.6 minutes (13.4 per minute)
At Poitiers University Hospital's emergency department, the X-ray discrepancy rate fell by 17% and mean stay by 9 minutes after Gleamer BoneView arrived (3,720 patients).
the rate of discrepancies between emergency physicians' and radiologists' readings fell by 17%
Teams' AI algorithms out-earned Dutch growers in Wageningen's greenhouse challenges; the 2018 winner raised sellable cucumber yield 6% and net profit 17%, and all five 2020 teams beat the growers.
increased commercially sellable production by 6% and net profit by 17%
The International Fund for Animal Welfare and Baidu report that AI Guardian screened 360,000 images and helped remove 7,853 illegal wildlife postings by mid-2024.
A convolutional neural network built at the University of California, Berkeley, found 72 missed pulses from fast radio burst FRB 121102; a 2023 search of part of the data re-detected 12 of them.
found 72 previously undetected pulses from the repeating fast radio burst FRB 121102
Facebook's machine-learning settlement maps cut the area that Red Cross volunteers had to map for Malawi's June 2017 measles vaccination campaign from 94,080 km² to 2,745 km².
reducing the area to map from 94,080 km² to 2,745 km²
In Las Vegas and Chicago in 2016 and 2017, 52.3% of 132 restaurants that Google's FINDER model flagged were unsafe on inspection, against 24.7% of 10,786 other inspections.
69 (52.3%) were deemed unsafe, against 2,662 of 10,786 (24.7%) baseline restaurants
Four researchers, two of them from the operator and the producer of the alerts, found that subscriptions to GLAD Landsat alerts cut the probability of deforestation in Africa by 18%.
decreased the probability of deforestation in Africa by 18 percent
In a blinded trial in Las Vegas from 2 January 2015, inspections of venues that nEmesis flagged from tweets found 9 demerits per inspection versus 6, and 11 venues graded C or worse versus 7.
found 11 venues graded C or worse versus 7 for controls
BioPatRec pattern recognition decoded a man's stump muscle signals for an augmented-reality limb, and his 48-year phantom limb pain eased to pain-free periods.
after ten weeks near-absent pain developing into completely pain-free periods
Authors including AIMS staff report that ReefScan surveys made island-scale reef assessments possible; an independent team found that ReefCloud's model annotations agreed poorly with a researcher's.
Pindrop's deepfake audio analysis, run against more than 120 voice engines, traced a fake Biden robocall before New Hampshire's January 2024 primary to ElevenLabs; the FCC later cited it.
the result came back well north of 99 percent for ElevenLabs
NICE reviewed 79 studies and concluded that AI autocontouring with clinician review, by systems such as MVision and Limbus Contour, is likely equivalent to manual contouring and quicker.
Reported time savings ranged from 3 to 80 minutes
Apple Watch sleep apnea notifications brought a 79-year-old man in Fukuoka to his cardiologists; his watch ECGs showed atrial fibrillation after nights rated “elevated”, and he began CPAP therapy.
The NAO reports that Justice Transcribe, an AI note tool, summarized over 1,600,000 probation meetings in England and Wales, an estimated 266,667 hours saved.
an estimated 266,667 hours saved
In university tests, Imagoworks' Dentbird software let technicians design crowns about four times faster, and Ohio State researchers concluded that its designs needed human review.
In May 2025 DeepMind reported that its Gemini-powered coding agent AlphaEvolve had found a 48-multiplication algorithm for 4x4 complex matrices, a rank that Igor Kaporin had reported in 2024.
4x4 complex-valued matrices with 48 scalar multiplications, improving on the 49
MegaDetector 5A removed 94% of images from manual review in a Florida snake study, but the hybrid workflow detected fewer snake events overall.
88,863 images remained for human review, a reported 94.0% reduction
In a UCLA Health randomized trial of 238 outpatient physicians, the Nabla ambient scribe cut time-in-note by 9.5% against usual care; DAX Copilot showed no significant change.
cut their time-in-note by 9.5% against the control group
In a 2024 University of Chicago pilot, 125 Abridge scribe users spent 15.9% less time in notes and 8.5% less time in the EHR than 478 weighted nonusers.
15.9% less time in notes and 8.5% less time in the EHR
In a 2024 trial with 259 science teachers in 68 schools, teachers told to use ChatGPT spent 56.2 minutes a week on lesson preparation against 81.5, with no evidence of lower quality.
saving of 25.3 minutes a week, or 31 percent
The Cour des comptes found that data mining, including machine learning in the French tax administration's CFVR program, came to originate half of tax controls and made choosing them more efficient.
originated half of tax controls but accounted for 13.8% of the tax and penalties recovered in 2023
Blind and low-vision users of Be My AI, the GPT-4 image-description feature of the Be My Eyes app, report in peer-reviewed studies that it read airport gate numbers and a handwritten check.
In a randomized trial with 60 Minnesota law students, GPT-4 cut time on four legal tasks by 11.8% to 32.1%, with slight and uneven gains in quality.
cut time on four legal tasks by 11.8% to 32.1%
In a preregistered experiment with 453 college-educated professionals, ChatGPT cut time on writing tasks by 40% and raised graded quality by 18%.
cut time on writing tasks by 40% and raised graded quality by 18%
Three of Udbetaling Danmark's machine-learning fraud models flagged cases that ended with benefits stopped or reduced 135, 36 and 72 times in 2023; Amnesty International questions their effectiveness.
135 cases with revenue (benefits stopped or reduced)
MutCompute, a structure-based machine-learning model, predicted stabilizing mutations for FAST-PETase, which almost completely degraded 51 post-consumer PET products in a week in laboratory tests.
degraded 51 post-consumer PET products in a week
Brazil's audit court (TCU) says that Alice, its screening robot, captured R$426m in 2021 (R$194m in 2020); an independent researcher reports that Alice helped auditors stop irregular tenders.
captured R$426m in 2021 (R$194m in 2020)
An 80-year-old woman's Apple Watch Series 3 showed a sudden drop in heart rate from 99 to 65 beats per minute at 19:29, when a Holter monitor recorded her reversion to sinus rhythm.
a sudden drop in heart rate from 99 to 65 beats per minute at 19:29
In Yokohama in December 2019, the AI Navi app cut taxi drivers' cruising time by 5% on average and 7% for low-skilled drivers, narrowing the skill gap by 13.4%.
cut taxi drivers' cruising time by 5% on average and 7% for low-skilled drivers
LendingClub's rating grades, which relied more and more on alternative data, moved some borrowers whom traditional criteria call subprime into better grades with lower-priced credit.
about 8 percent of A-rated and 28 percent of B-rated borrowers had FICO scores below 680
The Centers for Medicare and Medicaid Services count $654.8 million in inappropriate Medicare payments identified or prevented in 2015 with help from their Fraud Prevention System.
$654.8 million in inappropriate Medicare payments identified or prevented in 2015
Anthropic's safety monitoring flagged a chat user's threat to shoot up a Florida sheriff's office, and after human review the tip led deputies to detain her; no weapon was seized.
GPT-5.5 Pro solved Erdős problem #38; on 1 May 2026 the Aristotle system formalized the proof in Lean, and erdosproblems.com records the problem as solved.
Apple Crash Detection on a passenger's iPhone called the fire service about 46 minutes before the operator reported a sightseeing helicopter missing near Mount Aso; all three aboard died.
called the fire service about 46 minutes before the operator reported
After an alert from UNEP's Methane Alert and Response System, which uses AI, an operator in Kazakhstan rerouted gas to stop a pipeline leak; UNEP observed new emissions at the site by 21 May 2026.
estimated emissions of 1,200 tonnes
Sheng Siong's facial recognition system matched a wine thief to six prior thefts and flagged her return, which led to her arrest, restitution and an eight-day jail term.
matched a wine thief to six prior thefts and flagged her return
ChatGPT told Holli of Wolfforth, Texas, to go to the emergency room with a worsening spider bite; she says that staff there told her that the skin tissue was dying or dead.
Erken Uyarı Merkezi, an AI-based app built by Karadeniz Technical University students, alerted their phones about 30 seconds before a quake was felt at Turkey's parliament.
notifications had reached their phones 30 seconds before the quake
Google's Android Earthquake Alerts System warned Android users in Crete and the Dodecanese seconds before a magnitude 6.0 quake near Kasos on 14 May 2025.
The AI-supported Electronic Babylonian Library platform at LMU Munich matched tablet fragments, which let two scholars edit a lost Hymn to Babylon from 21 manuscripts in 31 fragments.
without the eBL platform the edition could not have used 21 manuscripts in 31 fragments
FamilySearch Full-Text Search converts handwritten records into searchable text, and a genealogist used it to help transcribe a 1761 deed; Transkribus claims 200 million pages deciphered.
200 million pages deciphered
In an NFER trial in 68 schools, teachers using ChatGPT with a guide spent 56.2 minutes a week on Year 7 and 8 science lesson preparation against 81.5 minutes for teachers without it.
56.2 minutes a week on Year 7 and 8 science lesson preparation against 81.5 minutes
The HALO Trust's AI detects mines in drone imagery of Ukrainian minefields with about 70% accuracy, and HALO says that it cuts analysis from days to hours; no cleared land is attributed to it.
achieving around 70% accuracy already
Kingston reports that Magic Notes cuts documentation time by 50–60%, with supervision write-ups falling from 40 minutes to under 10.
Comparing faces, a University of Bradford deep-learning model found 97% facial similarity with the Sistine Madonna in the de Brecy Tondo; Cartwright Hall showed it, and scholars called it a copy.
97% facial similarity with the Sistine Madonna
YES-TECH feeds satellite and weather data to a machine-learning model whose yield estimates set crop insurance claims in Madhya Pradesh; some farmers report faster payment and others lower claims.
kharif payouts in the three districts fell from Rs 217.91 crore to Rs 55.2 crore
Millie Houlton, 22, asked ChatGPT to draft an appeal against a wrongly issued £60 parking fine; York City Council withdrew it.
£60 parking fine; York City Council withdrew it
From 21 February 2023 the Supreme Court of India transcribed Constitution Bench hearings live with an AI service from TERES, and by February 2024 transcripts appeared on its website within days.
regularly published on the court website within days of a Constitution Bench hearing
A TrailGuard AI camera in Kishanpur Wildlife Sanctuary sent officials images of poachers with a gun and knives on two dates in 2022; officials identified them, and later sources claim arrests.
TrailGuard AI cameras sent tiger alerts to researchers and forest officials within 30 to 42 seconds, and officials later warned village leaders by text chat, though not rapidly.
received tiger notifications 30 to 42 seconds after detection
GiveDirectly's MobileAid model scored phone records to choose 3,052 of 12,800 recipients in a Malawi pilot, and the hybrid approach halved the cost per added enrollee compared with in-person surveys.
the hybrid approach cost about half as much per added enrollee as in-person surveys alone
The Beethoven X AI model generated continuations of Beethoven's sketches, which humans arranged into a Tenth Symphony that the Beethoven Orchester Bonn premiered on 9 October 2021.
A colorization model from the Google Arts & Culture Lab proposed colors for photographs of Klimt's destroyed Faculty Paintings, which the Belvedere showed at original size in 2025.
The Rijksmuseum used neural networks to reconstruct the strips cut from The Night Watch in 1715, and it displayed the painting complete from 23 June 2021.
a reconstruction of 393.1 by 507.4 centimeters
Groningen's AI writer-identification method split the Great Isaiah Scroll's 54 columns into two groups, which its authors attribute to two scribes; researchers in Norway questioned it in 2025.
two statistically distinct groups, with a transition in columns 27–29
Microsoft Translator added neural machine translation for Inuktut dialects with the Government of Nunavut; a Nunavik translator finds such tools helpful, and officials report errors.
At tax time 2020, nearest-neighbor checks in the Australian Taxation Office's myTax prompted nearly 340,000 users to review a claim; the office puts the revenue impact at about $37 million.
adjustments estimated to have a revenue impact of around $37 million
Rainforest Connection's Guardian sensors use AI to detect chainsaws and alert patrols; in West Sumatra an alert led a patrol to a logger in August 2019, and a patroller said that logging stopped.
Blackpool Council credits Project Amber, which detects potholes in road imagery with AI, with 5,145 repairs for just under £450,000 and claim payouts cut to £719; complaints later nearly doubled.
5,145 potholes were filled in North Shore for just under £450,000
Kyukyu VoiceTra, a speech-translation app for ambulance crews, reached 690 of Japan's 720 fire headquarters by 1 January 2025; more than 70% of Sapporo crews said that it sped up communication.
690 of Japan's 720 fire headquarters (95.8%) used the app
In a Virginia pilot, Crosstek software detected wildlife in thermal camera feeds with 97 percent accuracy, and drivers slowed by an average of 3 mph at the warning signs that it activated.
an average speed reduction of 3 mph as vehicles approached activated signs
Staff in Representative Moulton’s office report easier workflows with Civic Revere, but the interview does not separate AI from conventional software improvements.
Peterborough City Council reports that staff find emails and documents faster with OSCAR Workforce, a Datnexa AI tool; it gives no figures, and resident benefits remain intended.
At an emergency service in the Algarve, Lunit INSIGHT CXR flagged bilateral pleural effusion on a chest radiograph; ultrasound confirmed a right subpulmonic effusion, and the patient refused transfer.
Milton Keynes City Council reports that receipt-to-validation time for planning applications fell from 15.8 to 7.6 days while its officers piloted the Valon AI planning platform.
fell from 15.8 to 7.6 days
TU Munich researchers compared BirdNET with two experts on 15.5 hours of rooftop recordings; with tuned settings BirdNET reached an F1 score of 0.84 and found one species the experts missed.
BirdNET came close to the experts (F1 0.84)
Northeastern University Library linked Claude to its collections through a Model Context Protocol server and reported a better discovery experience in early testing; no source measures learning gains.
A nephew in Baku who is a dentist used ChatGPT, later GPT-5, to sum up studies and weigh off-label add-ons for a 65-year-old on FOLFIRINOX; 11 liver metastases regressed, cause unattributed.
11 liver metastases regressed
Zhejiang Lab reports that its GeoGPT platform identifies deep-sea species with 81.3% accuracy and takes 40 milliseconds per image, compared with 40 seconds manually.
Hubei Telecom reports that its Di Ting assistant shortened training for new agents from three months to one week and cut average handling time by 10%.
AgileSoDA reports that its reinforcement-learning physical-design system cut floor-planning time in partner deployments by 76%, from more than a week to roughly a day.
Huawei reports that its Pangu AI models made GAC's research and development work 30 percent more efficient; the joint submission presents other manufacturing gains as potential.
The UAE Ministry of Finance reports that MOF IntelliServe, a chatbot built on Azure OpenAI, gives instant bilingual answers and document search and reduces routine manual effort.
The UAE Ministry of Finance reports that AI Payroll Guard, which combines Llama 3.1 (8B) with robotic process automation, cut payroll processing time by up to 95 percent.
Dubai’s Smart School Finder pilot answers parents’ plain-language questions with KHDA school data; the makers say that it saves time and report no measured saving.
China Unicom's Yuanjing platform uses AI for dispatch, monitoring and event management, and the company reports a 25% improvement in urban management efficiency.
Shanghai AI Research Institute says that AI-assisted robotic parking runs in operating garages and estimates that single-layer layouts can raise capacity by 60–80%.
Changan reports that its Intelligent In-Vehicle Scenario Orchestration Agent cuts scenario setup from minutes to seconds; it reports no timing study.
Mobile AI sound cameras of Taoyuan's environmental department located a loud white BMW within a day of residents' complaint, and officers stopped it on 23 April 2025.
Elice’s AI Helpy gives tailored feedback in Korean AI digital textbooks, which lost their textbook status in August 2025; no source measures a learning gain.
Machine-learning methods run on single-specimen transcriptomes, including a variational autoencoder and CLADES, separated the Crozet population, which the authors named Limnognathia desmeti.
Huawei reports that RuiPath reduced pathology-review time at Ruijin Hospital, including initial review from 20–40 to 10–15 minutes per patient.
Birmingham City Council reports that its Foundry tools, built with Microsoft Copilot and UiPath, redact a document in 90 seconds, compared with 10 minutes by hand, and index 55% of emails.
Surrey Heath Borough Council reports promising early tests of Microsoft Copilot for meeting recaps and email summaries with two licenses, with no quantified gains.
Swindon Borough Council reports that Simply Readable, its tool built on Amazon Bedrock, makes Easy Read documents in under 30 minutes for 7–10 pence each.
The Mohammed Bin Rashid Housing Establishment in Dubai reports that AI credit and eligibility models automate housing-loan assessment, but no source measures their separate effect.
Greater Cambridge Shared Planning reports that the PlanAI language model cut a consultation-summary task from 18.5 officer hours to 16 minutes while officers still read every response.
Peterborough City Council reports that Hey Geraldine, an AI assistant modeled on an experienced occupational therapist, saves the therapy team 15 minutes per conversation.
ZTE reports that Jiupai, the hydrology model that it released with the Changjiang Hydrological Bureau, cut the time spent writing reports by approximately 30%.
A 62-year-old woman with retinitis pigmentosa used the Microsoft app Seeing AI, which turns printed material into auditory output, to access printed recipes in her kitchen.
Dubai clinic case: Oura ring heart-rate and HRV summaries, read by clinicians, fed into a stress test and Holter showing 2,078 PVCs in 24 h; the ring detects no arrhythmia and issued no alert.
24-hour Holter recorded 2,078 ventricular events (1.75% of beats, no couplets or runs)
Halton Borough Council reports that its Beebot AI family-hub platform offers information at any hour and that its automation has begun to reduce staff administrative work.
Ant Group reports that its medical AI assistant, later the AQ app, reached over 13 million users and that its annotation system improved efficiency by 50%; no health effect is reported.
over 13 million users
In a two-week diary study, 16 blind and low-vision adults used Eyevisor, an app built on Microsoft Azure AI Vision, and rated its scene descriptions 2.76 out of 5 for satisfaction.
2.76 out of 5 for satisfaction and 2.43 out of 4 for trust on average
North Yorkshire Council reports that its Microsoft Azure AI proof of concept for children's social-care records cut the time and cost of some data-retrieval tasks by 90% in testing.
90% less time and cost
West Berkshire Council reports that its job-advert generator, built on OpenAI GPT models, reduced managers' drafting time and made adverts more consistent.
ESTsoft reports that older users of its AI Senior Care Content and conversational companions in Gimhae senior centers described physical, psychological and social satisfaction.
Huadian and Huawei report that power forecasts from a fine-tuned Huawei Pangu weather model reduced assessment penalties by 27% and 15% at a Jiangsu pilot in the first half of 2024.
reduced assessment penalties by 27% and 15%
Barnsley reports that Copilot saves time on meeting notes, summaries, procurement, legal work, and social-care administration.
South Cambridgeshire District Council reports that its EBI.AI chatbot cut calls and improved routing; councillors later said that some callers hung up without help.
BattlegroundAI reports that its AI ad copy cut a school-board campaign’s cost per click by 5% and that combined human and AI copy had a 3% higher click-through rate.
cut cost per impression by 9% and cost per click by 5%
DeepMind's GNoME graph networks proposed 2.2 million crystal structures (380,000 stable); independent researchers found duplicates and scant evidence of useful new compounds.
2.2 million crystal structures (380,000 stable)
DP World reports that facial recognition in its Tanbeeh platform cut truck entry at the gate from three minutes to 15 seconds; prevented injuries are unproven.
West Oxfordshire and Cotswold councils report that AI tagging and summaries in Konveio and Go Vocal cut planning officers' repetitive administrative time by an estimated 85%.
The acting director of Dubai Central Laboratory said that its robotic X-ray fluorescence (XRF) system cut cement tests from four days to eight minutes.
Mathpresso reports that its QANDA app gives students real-time question solving, adaptive feedback and study plans; no source measures learning gains.
John Deere's See & Spray machine-vision sprayer gave weed control within 1% of broadcast spraying in soybean trials, with 28.4% early-postemergence and 36.5% mid-postemergence herbicide savings.
28.4% early-postemergence and 36.5% mid-postemergence herbicide savings
Victor, the AI project of Brazil's Supreme Court and the University of Brasília, has converted appeals to text since December 2020, with more than 10 million pages processed by May 2021.
more than 10 million pages processed by May 2021
Kanazawa pilot of 23 COVID-19 patients (2020): Fitbit oxygen-variation signals preceded 4 of 5 exacerbations (sensitivity 80%); one man's persistent signals led to a finding of severe sleep apnea.
EOV signals preceded four of five analyzed cases (sensitivity 80%, specificity 90%)
In FDA's review of ArteraAI Prostate, among 886 patients, 10-year metastasis risk was 28.1% in its high risk category and 3.3% in its low category, against 8.1% overall.
10-year metastasis risk was 28.1% in its high risk category
In FDA's review of Clairity Allix5, the software predicted five-year breast cancer risk from 77,511 screening mammograms with an AUC of 0.70, above a goal of 0.64.
an AUC of 0.70, above a goal of 0.64
In FDA's review of the Prenosis Sepsis ImmunoScore, sepsis within 24 hours rose from 3.02% in its low risk category to 69.70% in its very high category; AUROC was 0.81.
an area under the ROC curve of 0.81
In FDA's review of Viz HCM, the software flagged hypertrophic cardiomyopathy on 12-lead ECGs from three hospitals with 68.4% sensitivity and 99.1% specificity.
a sensitivity of 68.4% and a specificity of 99.1%
In the 14-site study that FDA reviewed, the Cognoa ASD Diagnosis Aid gave a result for 32% of 425 children, with 81% PPV and 98% sensitivity against specialist diagnosis.
a positive predictive value of 81%
In a retrospective test that FDA reviewed, Viz.ai ContaCT sent its stroke alert in 7.32 minutes on average, against 58.72 minutes for usual-care notification of the specialist.
a mean difference of 51.40 minutes
A retrospective evaluation suggests that Google Flood Hub forecasts could extend some early-action plans in Mali from four to seven days, with frequent false alarms; no early action was measured.
some early-action plans could extend from four to seven days without much loss of forecast skill
In a blinded study at a German pig nursery, the SoundTalks cough monitor's respiratory health score fell with high swine influenza A loads; no intervention outcome is reported.
In a Nature Communications study, Level 4 automated driving systems from unnamed makers had lower crash odds than human drivers in most matched scenarios and higher odds at dawn, dusk and in turns.
In rain, a Level 4 automated vehicle had 0.335 times the crash odds of a human-driven vehicle
Independent researchers at Colorado State University found that DeepMind's GraphCast and three other AI weather models matched the best operational models on 2023 cyclone tracks.
adding all four improved the NHC track consensus by 11% at 120 h
DeepMind says that a live GraphCast run predicted about nine days in advance that Hurricane Lee would make landfall in Nova Scotia; no source documents a forecaster using the run.
about nine days in advance that Hurricane Lee would make landfall in Nova Scotia
The IRS Large Partnership Compliance Model, an AI model, flagged 150 large partnership returns for review in April 2023; no examination that it selected had closed by May 2025.
DeepMind's Ithaca, a deep neural network, raised historians' accuracy at restoring damaged ancient Greek inscriptions from 25 to 72 percent in a test; no source reports an accepted restoration.
raised historians' accuracy from 25 to 72 percent when they worked with it
Colombia's Constitutional Court presented PretorIA on 27 July 2020 to classify and search tutela rulings for selection, and the court still used it in 2024; no source measures an outcome.
A directed-message-passing neural network built at MIT identified halicin, which treated infections in mice; independent laboratories confirmed its activity, and no human trial is reported.
Transport for NSW announced a trial of its Protective Animal Warning System, in which two AI-enabled cameras on the Monaro Highway would trigger driver warnings; no results are reported.
Odakyu's AI crossing image-analysis system reads existing crossing cameras and signals approaching trains to stop when someone is left inside; trains made 17 emergency stops in two months.
trains made 17 emergency stops on detected anomalies in the system's first two months
Qualcomm says that a humanoid on its Dragonwing IQ10 platform ran a designed “safe-collapse” sequence when it fell at Computex 2026; news reports describe a face-first or backward fall.
WhaleSpotter software flags gray whales in thermal camera images for a specialist to verify before alerts reach San Francisco Bay mariners, and detections had passed 16,900 by early July.
detections had passed 16,900 by early July
Greece's deputy transport minister described an AI system from an unnamed supplier that transcribes railway radio calls and checks them against protocols; he gave no measured safety result.
Troopers said that the eye detection feature of a Tesla on Autopilot halted the car on I-75 after its intoxicated driver fell asleep; no vehicle data confirm the sequence.
Lime Vision, Lime's in-house camera AI, detects sidewalk riding and sounds repeating alerts; Seattle's transport department reported its rollout on 9 June 2026 without outcome data.
Footage of 2 March 2026 may show Israel's Iron Beam laser downing a Hezbollah drone; the military said on 12 March that Iron Beam had not been used, and only the OECD entry calls the laser AI.
one drone intercepted near Shlomi
ProRail replaces 288 lineside cameras with smart cameras in 2026 and runs Genetec video analytics on its 7,000 cameras to flag trespassers to dispatchers; no results are given.
In a preprint on 96 patients, CA-GPT, built on DeepSeek-R1, agreed with expert stent plans more often than junior physicians and ChatGPT-5, which received only extracted measurements.
stent diameter (90.3% vs. 72.2%, P<0.05) and length selection (80.6% vs. 52.8%, P<0.01)
MuleHunter.AI, from the Reserve Bank Innovation Hub, flags likely mule accounts at 26 Indian banks; the Reserve Bank of India has withheld the number of accounts found.
At DEF CON 33 on 8 August 2025, DARPA reported that seven teams' cyber reasoning systems found 54 synthetic vulnerabilities, patched 43, and found 18 real bugs for disclosure.
found 54 synthetic vulnerabilities, patched 43, and found 18 real bugs for disclosure
The Department of Telecommunications' ASTR facial-recognition tool flags connections that one person took under different names; more than 82 lakh were disconnected after failed re-verification.
disconnected more than 82 lakh connections through ASTR after they failed re-verification
FDA reports that its imported-seafood model recommended samples at 328 U.S. ports and that its Elsa tool helped find inspection targets, with no figure on hazards found or illnesses prevented.
A relative credited ChatGPT's paraffin-oil advice with saving a teenager from a grain pill, but the medical dean at Assiut University said that the teenager showed no signs of poisoning.
Commentary credits Rafael's X-Guard towed decoy, labeled “AI-powered”, with diverting Pakistani missiles from Indian Rafales; no official source confirms its use, and India admitted jet losses.
Rete Ferroviaria Italiana's AlpEW seismic early-warning system reportedly stopped Rome–Naples trains during a magnitude 4.4 earthquake; its developers describe no learned AI model.
a magnitude 4.4 earthquake, during which trains on the Rome-Naples line were stopped
In laboratory tests on real and synthetic CT scans, Rapiscan's AI algorithms detected shark fins, seahorses and sea cucumbers at 86–96%, with 13% overall false alarms; no customs use is reported.
detection rates of 95% for shark fins, 96% for seahorses, and 86% for sea cucumbers
NBC News reported that Lynn White used ChatGPT and Perplexity to overturn her eviction on appeal; court records show that her later petitions in the case were denied in November 2025.
avoided roughly $55,000 in penalties and over $18,000 in overdue rent
The Kenya Wildlife Service installed FLIR thermal cameras whose built-in AI alerts rangers to intruders at three black rhino sites; no arrest or prevented poaching is linked to them.
DeepMind's AlphaProof and AlphaGeometry 2 were awarded 28 of 42 points, a silver-medal score, on the 2024 International Mathematical Olympiad problems, with some solutions taking up to three days.
awarded 28 of 42 points, a silver-medal score
Kota police receive Meta's red-flag tags for suicide and self-harm posts on Facebook and Instagram; within a week, police said that they had prevented a possible suicide by a Jhunjhunu student.
Convolutional translation models by Tel Aviv University and Ariel University researchers rendered Akkadian into English with BLEU4 scores of 36.52 and 37.47, best on short sentences.
BLEU4 scores of 36.52 and 37.47
An AI monitoring system from OKI and Marubeni Network Solutions and a 3D stereo-camera system detect people left in Seibu Railway crossings and light lamps that warn train drivers.
IdentiFlight V5 classified lesser spotted eagles correctly in 97.8% of checked images at a German test site in 2025; no German source measures a drop in bird deaths.
Ve.R.A., an algorithm of Italy's tax agency, cross-matches financial and tax data under pseudonyms to flag taxpayers at higher risk of evasion; no source reports a result.
Meta's platform flagged a Paraguayan teenager's post praising the Uvalde shooter, and after the FBI relayed the alert, police, prosecutors and psychologists visited him.
ScamShield, a Singapore government app whose on-device machine-learning model filters text messages, flagged about 5 million of them to users as potential scams from March 2022 to November 2023.
about 5 million text messages had been flagged to users as potential scams
Google's Android Earthquake Alerts System warned Android users near Alanya, Turkey, seconds before a magnitude 5.1 offshore quake on 5 January 2022.
Hürriyet put the lead at about 10 seconds
In July 2021 Sanyo Electric Railway began to use Optage's AI analysis of crossing camera video to detect people left inside two crossings and warn train drivers.
Houston Audubon issues Lights Out alerts when BirdCast's model forecasts high migration, and Houston proclaimed Lights Out Nights on 9 April 2021; no Houston collision reduction is documented.
The Nanjing Building's digital twin on Huawei Cloud relayed a window sensor's alert to a guard's wristband, in an anecdote that the sources place on three different holidays.
995 door/window alerts from January to May 2021
Kuzi, launched by the Nairobi firm Selina Wamucii in January 2021, uses machine learning on satellite, soil-sensor and weather data to map locust risk; no source reports an alert that helped.
Whale Safe, launched on 17 September 2020, uses automated whale-call detection and other data to rate whale presence daily and to grade ships' cooperation with voluntary 10-knot zones.
an 85 percent cooperation rate overall by fleets
Nuru, a machine-learning app from Penn State's PlantVillage and FAO, launched on 25 June 2018 to confirm fall armyworm damage offline on Android phones; no source reports an outcome for farmers.
Delhi Police's facial recognition, which The Wire identifies as Innefu Labs' AI Vision, produced 2,930 matches among children in care, and the High Court later recorded that it cracked no case.
produced 2,930 matches among children in care
Raleigh police detectives used Traffic Jam, a Marinus Analytics tool that searches escort ads by phone number and face, to corroborate victims' accounts of a trafficker's movements.
In a double-blind retrodiction of 1,637 Chicago food inspections from 2014, the city's Food Inspection Forecasting model would have found critical violations on average 7.5 days earlier.
would have found 178 (69%) in the first half of the work, on average 7.5 days earlier
North East Derbyshire District Council uses Microsoft Copilot for drafting, summaries and analysis and expects it to cut repetitive work; it reports no measured saving.
Researchers used MADGRAV-2D to recover 47 cataloged gravitational-wave signals from LIGO data.
47 detections with a false alarm rate below one per year and p_astro above 0.9
Network Rail reported that trespass at Dingwall fell after it installed DeterTech AI cameras alongside signage and campaigns; it gave no later count and noted a renewed rise.
The Metropolitan Police said on 13 April 2026 that it is testing unnamed companies' AI tools to grade abuse imagery and flag unknown victims; no use on live cases is reported.
Open Data Labs says that an unnamed retailer's undisclosed model, trained on consented data, predicted style preferences better than baseline methods; it gives no accuracy values.
GrantRail, built with Airtable Omni, demonstrates AI support for education funders on synthetic data; its maker says that one prompt produced a report in four minutes, and no user is named.
Models built from MegaDetector and SpeciesNet labels came close to expert-based models, with mean occupancy differences of 0.03-0.07 across study areas; some species were missed or misclassified.
Mean occupancy differences of 0.03-0.07 across study areas
Metalesa's PlugSmart Pro system uses cameras and detection algorithms to spot wild boar near the CV-600 road and light warning signs only when animals are present.
South Korea's transport ministry plans AI cameras from an unnamed supplier to detect vehicles or pedestrians trapped at crossings and alert drivers, with pilots at Nonsan and Boseong.
In post-processing, the CardioWatch photoplethysmography algorithm detected cardiac arrest at 09:50:17 after loss of pulse in a cyclist; his implanted defibrillator ended it, and no alert was sent.
detected cardiac arrest at 09:50:17 after loss of pulse
Run afterwards on light-sensor data from a Samsung Watch5 Pro worn by a dying patient at Amsterdam UMC, the HEART-SAFE algorithm marked circulatory arrest in step with the electrocardiogram.
Cambium's Chat with Personas lets teams ask synthetic personas, built from public data, about prices, messages and product ideas; Cambium reports no measured gain in research.
A Manchester residents' panel proposed safeguards and a fourth design for a public-facing version of Caddy, an AI advice tool from Citizens Advice; no source reports that the service launched.
Yourcore’s founder describes AI-assisted app development and an intended style companion; benefits to users’ wellbeing or consumption remain unmeasured.
VoteMate’s free AI ballot assistant answers voters’ questions about candidates and ballot measures; tens of thousands used it before the 2026 midterms, and no source measures what they learned.
Given slit-lamp photos, ChatGPT (GPT-5) named oncocytoma first in two retrospective cases and gave a wrong leading diagnosis in three exploratory tests; patient care was not affected.
WINGS says that its wi.breathe cameras and AI smoke and fire detector alerted authorities to a Troizina wildfire on 28 August 2025; the municipality's statement that day does not mention them.
Mines Eye, an aerial system whose AI detects explosive remnants of war, scanned over 360 hectares in a five-month pilot by Ukraine's emergency service; the release gives no detection counts.
scan over 360 hectares of confirmed and suspected hazardous areas
Northumberland County Council's FloodAI machine-learning service aims to forecast flash floods in six catchments; it was still a prototype in October 2025.
C-DAC Pune reports that its AI-based Virtual Yoga Instructor, a MoveNet-based pose classifier, identified ten yoga poses with an accuracy of 0.86 on test data; no practitioner result is reported.
CERN reports that its CAFEIN federated learning platform is in operational use; in a test on a public stroke dataset, federated training outperformed training on local data alone.
CSIR-INSTI and Bolgatanga Technical University propose a smartphone-over-microscope TinyML model to help health workers detect malaria parasites; they have published no field test or accuracy result.
iSciLab describes an AI system, including its AniPuppy app, that identifies animals by nose prints and facial features; it lists Korean and African pilots and reports no outcome.
Tata Elxsi describes an early AI system for remote patient monitoring and triage support in underserved communities; no patient, clinic or health system is reported to have used it.
NT Health plans a Bayesian hierarchical model and Power BI AI features to guide ear-health services; by 25 May 2026 the mapping phase was complete and the AI tools were still being selected.
AIforPet's TTcare Equine identifies horses and analyzes their gait; AIforPet reports that pilots showed feasibility and user acceptance, without figures.
Christ University, Bangalore, is developing MoralQuest, a game of interactive moral tales that is meant to personalize learning and social development for autistic children.
Anna University proposes a system in which a camera, a Raspberry Pi and a cloud model score rehabilitation exercises and give live feedback; testing is planned.
Emirates Health Services runs Ask EHS, an AI chatbot, alongside multilingual and accessibility tools on its website; no source reports a measured change in access or workload.
AI4Africa proposes an AI tool that converts hard-to-understand speech into clearer speech and plans to seek Limi Hospital or another Nigerian hospital to test it.
Mbeya University of Science and Technology proposes a wearable AI-Based Health Monitoring and Advisory System that would send alerts and advice when readings are abnormal.
EquiDermAI uses generated skin-lesion images and compact on-device models to narrow skin-tone gaps in diagnosis, and its makers report that full validation is ongoing.
StudentSphere is developing an AI journaling tool that would give students emotional feedback, and an internal pilot at VIT is planned.
VIT researchers trained brain-wave (EEG) models on data from ten people and plan hospital tests of a wheelchair that its user steers with brain signals.
VIT researchers describe a smartwatch system for monitoring facial paralysis and plan to collect a dataset and build a model that recognizes facial exercises.
The makers of the AI-Enhanced Organoid Platform aim to cut drug-screening time by 50%, citing other teams' studies; from partner tests they report only latency and user satisfaction.
data354 plans AI maps of mangrove biomass and carbon in Côte d’Ivoire; its October 2025 paper derives biomass from 300 field plots and lists the maps as future work.
AIrth has tested an energy-insight prototype with simulated data and plans pilots with Georgian businesses and households.
TeleAI describes learned scheduling with large language model explanations for computing clusters; its reported GPU gains come from virtualization and a heuristic scheduler.
SBER describes an object detector and a classifier for 25 species in camera-trap images, meant to cut manual image review; a public demo ran in 2025, with no reported field result.
Ishango's Project Umoya plans to fine-tune GraphCast and Aurora on African data; in a partner master's study, a fine-tuned small Aurora beat pretrained Aurora on past South African weather.
Tata Consultancy Services proposes a system that uses semantic scene understanding and generative rendering so that disaster responders can share visual information over limited bandwidth.
Vellore Institute of Technology is developing an AI system to forecast low-oxygen ocean dead zones two to four weeks ahead, with regional pilots planned.
Mbeya University of Science and Technology in Tanzania proposes machine-learning analysis of radio signals to locate people during search-and-rescue operations.
Christ University, Bangalore, is developing an AI-enabled urban heat advisor that would identify vulnerable areas and support targeted cooling and energy decisions.
Sri Sivasubramaniya Nadar College of Engineering is developing a Vision Transformer model to predict local microclimates for urban design, heat-risk management and irrigation.
Mbeya University of Science and Technology plans a machine-learning module that would predict elephant conflicts from movement logs; its current prototype uses GPS and fixed geofences.
The University of Sadat City and SAMI Advanced Electronics are developing an AI model to forecast wind on the Red Sea coast and the Gulf of Suez; operational validation is still planned.
The Ethiopian Red Cross and the Ethiopian Meteorological Institute are partners in a proposed AI chatbot that would give multilingual disaster guidance; pilots are planned.
China Telecom HongXin describes its AI Full-Process Low-Code Middle Platform, an internal tool for building AI applications in transport, governance, and environmental monitoring.
China Telecom describes an internal pilot of its network large model toolchain platform, which is meant to streamline cloud network operations and collaborative development.
Tata Elxsi is testing AI-driven telecom digital twins in simulated environments to identify energy-saving configurations and reports no measured saving.
The Venus team at Shanghai University proposes an AI application that predicts ionospheric delay in real time and reports no measured gain in positioning or communication.
HFCL proposes AI-based radio-frequency sensing that detects and classifies objects for telecom, transport, manufacturing, and monitoring, and reports no testbed result.
IIT Gandhinagar researchers propose learned traffic forecasts and graph neural networks with deep reinforcement learning for adaptive routing and report no simulation result.
IIT Hyderabad researchers plan to switch 5G network-slice instances as load changes, with machine learning models trained on synthetic data, and report no testbed result.
Addis Ababa Science and Technology University proposes AI-driven predictive beamforming for 5G and 6G links in bad weather, with simulation and an open-source toolkit still planned.
An Indian Institute of Technology Hyderabad team reports that its cross-slice federated learning model reached a macro F1 score of 93.83% on a test dataset with new anomalous traffic.
Amrita University proposes a network digital twin that combines geographic data with AI-supported radio adjustments to run 5G and 6G base stations on less energy; no measured result is reported.
Tata Consultancy Services built a proof of concept that uses semantic communication and generative AI for low-bandwidth interaction with a virtual 3D demonstrator; no learning outcome is reported.
MUST simulated a free-space optical link in MATLAB while training an AI model to predict attenuation from climate data; its July 2025 code holds laser-link sketches and no AI model.
The Emirati Ministry of Community Empowerment reports that its name check, which uses a fine-tuned Sentence-BERT model to compare Arabic and English nonprofit names, passed acceptance tests.
The Centre for Gender and Politics proposes AI features for multilingual access and mentorship matching in its GaP Repository; on 29 October 2026 the live site offered none of them.
The #BlackTechFutures Research Institute reports that HEART Engine uses recommendation models and conversational agents to tailor training and job matching but quantifies no gain.
Hangzhou Dianzi University reports that its content moderation system detected deepfakes with about 95% accuracy in a challenge and that public security bureaus in China apply it.
A Zambian final-year research project proposes an unnamed AI model that would summarize terms of service and privacy policies and flag clauses that could compromise users' rights.
UNHCR's Uncovering Delays project applies machine learning and process mining to historical case data to find delays in refugee status decisions; no time saving is reported.
UNHCR is building the RiMAP Virtual Legal Assistant to collect, summarize and analyze laws affecting displaced and stateless people; its 30 percent cut in research time is a target.
Mbeya University of Science and Technology pilots an AI-driven wireless audio conference system that uses smartphones as microphones, queues speakers and transcribes meetings.
Addis Ababa Science and Technology University proposes an edge-and-cloud AI intrusion-detection system for Internet of Things networks and has tested it only in a simulated deployment.
The Dubai Government Human Resources Department proposes an AI-powered talent-matching platform to connect job seekers with suitable roles and reduce hiring bias; no pilot is reported.
The Indian Institute of Information Technology Sonepat proposes a generative AI system for virtual try-on and garment design that is intended to cut returns and textile waste.
Sevnce Robotics reports pilots of its explosion-proof inspection robot at a substation and a chemical plant and states its efficiency and cost gains only as expectations.
BSD design is developing XTENDIT, AI-based extended-reality training in automation skills for industrial workers, and reports no learning or onboarding result.
Mbeya University of Science and Technology is developing AI models that assess concrete cracks and moisture from images and sensor data; no field result is reported.
Zamtel is piloting AI credit scoring and fraud detection over its mobile network; better access to loans, higher incomes and consumer protection remain intended benefits.
Two independent researchers propose a three-tier AI framework that verifies NGO registrations, monitors transactions and prioritizes audits; its 30% fraud reduction is a projection.
Receipt Guardian, an app meant to help children and families manage spending, is in beta with more than ten testers; better financial literacy and spending habits are only intended.
Researchers at the Vellore Institute of Technology propose a model that adjusts credit-scoring thresholds for subprime lending and have tested it in synthetic simulations.
Researchers at IIT Hyderabad are building InsurVision, a set of language-model agents meant to analyze insurance claims; faster settlements remain an intended benefit.
SERA proposes Smart Budget, an AI service that would personalize budgets and suggest spending reductions and income opportunities; no change in users’ finances is measured.
Al-Aqsa University is developing an offline multi-agent AI learning system that is expected to sustain teaching in crisis zones; no source reports a result.
Addis Ababa Science and Technology University plans a nine-month pilot of an AI instructor for a 3D electronics lab; no source reports a result.
Addis Ababa Science and Technology University proposes an adaptive AI learning assistant for neurodivergent students; no source reports that students used it.
Sign<Tech> is piloting Little Signs, a speech-to-sign-language tool, at a Zimbabwean school; no source reports a measured result in the classroom.
Rein4ce Learning, a Ugandan company, proposes an AI personalization framework for education, wealth, nutrition and health; no pilot or result is reported.
Culturalink organizes heritage information and generates cultural narratives, and it uses the DeepSeek language model to link heritages; no source measures a learning or preservation result.
iMUSIC describes an unnamed AI system that generates personalized video ringback tones, and its account reports no pilot or realized communication benefit.
Yi Jiang Jie, a museum guide built on DeepSeek-R1, gives visitors contextual explanations and personalized routes, and no source measures visitor or museum outcomes.
Mbeya University of Science and Technology in Tanzania proposes training an AI model on dish misalignment to speed up realignment; no test of the model has been reported.
AI4Africa proposes an unnamed AI system that forecasts connection quality for online events; its only testbed is a public Kaggle dataset, and no pilot is reported.
An unnamed AI layout optimizer from VIT University in India has been tested only on synthetic and reference floor plans; more usable space in low-income homes remains a goal.
Catalink built an AI drowsiness detector for drivers; in July 2026 its researchers reported that the models they tested generalized poorly and that phone versions fell to near chance-level accuracy.
A UNHCR team fine-tuned a Kurdish speech recognition model, whisper-small-Kurdish-Sorani-10, to transcribe humanitarian interviews; no use with interviews is reported.
The International Institute of Information Technology, Hyderabad, is building a proof-of-concept AI system to recognize and translate Indian Sign Language; no test with signers is reported.
A prototype from Addis Ababa Science and Technology University uses vision models and eye tracking for hands-free computer control; no test with people who have hand impairments is reported.
SUPSI and IDSIA built a laboratory demonstrator of the REXASI-PRO autonomous-wheelchair safety system; no source reports a result for wheelchair users.
TCS Research reports that its AI framework for decisions by vehicle groups predicts quality of service better for following vehicles; no source reports a road trial or a safety or fuel result.
OptM describes controlled road pilots of its AI vehicle-to-vehicle coordination system; the company has published no result on safety, traffic or emissions.
AI4Africa proposes an AI-enabled 6G-V2X communication framework for connected vehicles; the team reports that the work remains in training and simulation, with no results.
AI4Africa proposes a deep learning system that detects intrusions in vehicle-to-vehicle messages; no detected attack or test result has been reported.
A system that pairs Chronos forecasts with Mistral 7B advice improved energy efficiency by about 9.2% in simulations; it has not been deployed, so effects on real vehicles are unknown.
The CSIR-INSTI team built a TinyML weather-monitoring prototype for a proposed farm-advisory system and later field-tested a soil sensor in Ghana; no gain for farmers is reported.
Hurudza AI, a contact center that uses a GPT-2 model trained on Shona, gives farmers advice by voice and text in pilots in three wards of Zimbabwe; its gains are projections.
China Telecom reports that its Divine Farmer model answers farming questions with over 90% accuracy and recognizes pests and diseases with over 85% accuracy.
Mbeya University of Science and Technology pilot-tested a YOLOv9-based edge AI system that detects tomato diseases on its farm; no avoided crop loss, yield gain or pesticide saving is measured.
Mbeya University of Science and Technology is developing soil sensors and a random-forest model that predict suitable crops and fertilizer; no source reports use by farmers or any gain.
A Dubai Government Human Resources Department use case proposes an AI decision-support toolkit for human resources staff; the sources name no software and report no result.
The Indian Agricultural Research Institute proposes AI rainfall forecasts for its automatic basin-irrigation system to skip watering when rain is expected; their saving is unmeasured.
A Nara Knowledge Information patent describes AI speech recognition for a metaverse that blind users control by voice; no source shows that its Blind Metaverse demo uses that function.
Virginia piloted Vulcan Justinian to analyze state regulations in 2025 and ended the pilot in 2026; no source establishes how much the AI contributed to claimed savings.
A regression-based neural network from Universiti Putra Malaysia predicts how regional weather affects power-line data throughput in Malaysia and Nigeria; no source reports use by a utility.
Akimilabs is testing Etcera, a multi-agent AI assistant intended to reduce decision fatigue and cognitive overload, with selected users and has published no measured outcomes.
Open Climate Fix reports that predicted satellite images from its Cloudcasting model raised the accuracy of its PVNet solar forecasts by 5% at two hours ahead.
a 5% accuracy gain at a two-hour horizon
Warwickshire County Council's locally hosted Llama 2 model classified businesses with 63% accuracy, against 95% for a simpler method; no time saving or gain in food access is reported.
63% accuracy, against 95% for a simpler method
Newham Council and the University of East London planned an unnamed AI model to forecast housing demand and homelessness; a partner later said that the models inform allocations, giving no results.
Researchers synthesized TaCr2O6 proposed by MatterGen and estimated its bulk modulus at 169 GPa against a 200 GPa target; a chemist later found that the compound had been known since 1971.
estimated its bulk modulus at 169 GPa against a 200 GPa target
Mount Sinai authors report that ForeSight-HF (GPT-5 mini on ProviderLoop) generated diuresis advice for a warm/wet and de-escalation advice for a cold/dry nursing-home patient; neither guided care.
AInnovation integrated ChatCAD with Bentley software and released iPID, a product based on it, with Bentley in August 2025; iPID's efficiency gain of more than tenfold is a preliminary estimate.
In a phase 1 trial in 51 healthy adults, GB-0669, an antibody optimized with Generate Biomedicines' machine-learning models, was tolerated, had a 54-day half-life and neutralized virus in serum.
dose-dependent live-virus neutralization at 600 and 1,200 mg
Sum Vivas supplied multilingual AI digital humans to assist visitors at Liverpool's AI Summit on 21 November 2024; no source reports how many people they helped.
Chinese neurologists used ChatGPT-4o to help review published MRI findings of cerebral intravascular large B-cell lymphoma and state that its conclusions agree with those of human experts.
The UAE Ministry of Finance reports that an AI assistant on its website gives real-time guidance on FATCA and Common Reporting Standard (CRS) tax reporting.
Milton Keynes's 2024 AI festival offered a hands-on AI workshop and an experience center with interactive AI applications and robots, including Pepper; no source measures what visitors learned.
The HIBA chatbot, built on GPT-4, gives health information in text and audio; its makers report about 700 user queries in six months and use in shelters in Lebanon.
Austin Energy announced on 29 August 2024 that a 13-camera Pano AI wildfire detection system covers its service territory; no source reports a fire that the cameras detected.
EDIT by DIY Render, from 3D Architectural Solutions, generates and revises property images for design discussions; no source measures time saved or faster sales.
XR Guild's founder proposed an AI librarian for the Guild's ethics library, and one catalog entry credits ChatGPT with its keywords; no source reports that the librarian launched.
ECMWF ran Huawei's Pangu-Weather from its own analyses and found accuracy comparable to its IFS model, with better summer temperature-event forecasts in Europe.
Pangu-Weather discriminated events better than the IFS
The Federal Authority for Government Human Resources added an AI assistant to its FAHR Smart App for federal employees, and no source measures an outcome.
Fondation Hirondelle's Radiolytica transcribes French and Swahili community radio to detect hate speech in the eastern Democratic Republic of the Congo; no source reports a measured result.
Fondation Hirondelle reports that AI software, which no source names, transcribes ten radio stations in eastern Congo in Swahili and French so that keyword searches can flag hate speech.
UN Women’s AI School has participants use ChatGPT, Gemini and Microsoft Copilot and build custom GPTs for research and advocacy; no source measures what the AI changed.
The Low-Frequency Detection and Classification System on a glider detected right whales east of Portland, Maine, on 27 December 2023, and NOAA declared a voluntary Slow Zone.
route around the area or transit at 10 knots or less
At two Scottish salmon farms in 2023, Observe Technologies' neural network detected that fish activity rose from 25.6% to 43.6% (Farm A) and 24.9% to 32.6% (Farm B) after gill health deteriorated.
fish activity rose from 25.6% to 43.6% (Farm A) and 24.9% to 32.6% (Farm B)
ENEA and two Italian universities tested a machine-learning model that forecasts grid failures during heatwaves; ENEA's announcement of 18 May 2023 reports no avoided outage.
Researchers found that contributors using the AI-assisted RapiD editor mapped up to 12.4% slower after the 2023 Türkiye earthquake; the OpenStreetMap wiki reports that it was then disabled.
contributors using RapiD mapped tasks up to 12.4% slower
CapisciAMe, a speech recognizer tailored to a woman with aphasia and dysarthria, identified 72.65% of 936 recorded test words; twelve professional listeners averaged 56.75% on a separate task.
identified 680 of 936 test utterances (72.65%)
Trinity College Dublin's ABAIR team reports that its Irish synthetic voices power an NVDA screenreader plugin with Braille output, tested with visually impaired pupils; no outcome is reported.
Xuanwu Hospital implanted the Epilcure responsive neurostimulator in a man with epilepsy in March 2022; its trial team reports a 61.1% mean seizure reduction in seven patients and no result for him.
A proof of concept that Faculty built with NHS AI Lab Skunkworks in 2021 forecasts bed demand and suggests beds; Kettering General Hospital tested it, and no source reports live use.
UN plans for early typhoon aid in the Philippines have used 510's XGBoost model of house damage as the trigger since 2021; the only documented activation, in 2025, came from observed wind speed.
Dytective, a gamified online dyslexia test with a random-forest classifier, screened 3,644 children with 79.8% accuracy, and two 2026 papers found a demographic confound in its dataset.
79.8% accuracy with 80.6% recall and 79.3% precision
In a five-hospital study by the TREWS developers, patients whose sepsis alert was confirmed within three hours had 3.3% absolute lower in-hospital mortality, a link that outside researchers question.
3.3% absolute (18.7% relative) lower in-hospital mortality
Whale Safe rates whale presence from acoustic buoys, sightings and a habitat model and grades ship cooperation with California's voluntary 10-knot zones; no avoided strike is documented.
Researchers at icipe used MaxEnt, a machine-learning algorithm, to map likely desert locust breeding areas in Kenya, Uganda, South Sudan and Sudan; no source reports use in the field.
the Mauritania model had the highest mean AUC (0.887)
In a simulation of past near-miss cases at a Tokyo Bay traffic center, collision-risk predictions from Fujitsu Human Centric AI Zinrai helped controllers send warnings about two minutes earlier.
the number of warnings issued to risk-prone vessels nearly doubled
In a field evaluation in Kenya and Tanzania, the offline PlantVillage Nuru app identified cassava disease symptoms with 65% accuracy, above extension agents (40–58%) and farmers (18–31%).
65% accuracy in 2020 versus 40–58% for extension agents and 18–31% for farmers
Against expert microscopy, the EasyScan GO machine-learning microscope read 2,250 malaria blood films at 11 sites with 91.1% sensitivity and 75.6% specificity; no patient outcome was measured.
Diagnostic sensitivity was 91.1% (95% CI 88.9–92.7) and specificity 75.6% (73.1–78.0)
Indonesia seized the MV NIKA on 12 July 2019 after INTERPOL tracking aided by OceanMind and Global Fishing Watch; no source names AI software or documents an AI detection.
the captain was convicted, the owner fined Rp300,000,000 and the vessel forfeited
IIIT Bangalore presents a machine-learning pipeline that detects necrosis in meningioma images and intends it for clinical decision support.
Corti's machine-learning framework detected more cardiac arrests on Copenhagen emergency calls than dispatchers did, but its alerts did not significantly improve dispatcher recognition in a trial.
higher sensitivity than dispatchers (84.1% versus 72.5%)
OPTiM says that its AI Physical Security Service watches JR Kyushu station camera video in real time and alerts administrators to anomalies; no trial result has been published.
KYield's HumCat, built on the Kyield Operating System, is meant to detect organizational risks and prevent human-caused catastrophes; no source documents a catastrophe that it prevented.
A support-vector machine classifier from Karolinska Institutet identified 9–10-year-olds at risk of persistent reading difficulty from under one minute of eye tracking with 95.6% accuracy.
reached 95.6% (± 4.5%) accuracy with balanced sensitivity and specificity
A Darkriderdesign submission proposes that social workers use Copilot to combine records for people experiencing homelessness; no source reports that it was built or used.
StratLoom describes AI-assisted curriculum work and supervised prompt practice in its Digital Roots programs; it publishes no evidence that learning or educator efficiency improved.
No cases match.