{ "name": "AI on the Record: sourced ledger of measurable AI results", "url": "https://ai-on-the-record.pages.dev/the-record/", "license": "CC BY 4.0", "checked": "2026-09-06", "count": 16, "claims": [ { "id": "masai-detection", "area": "Medicine", "headline": "AI-supported mammography found 29% more cancers with no increase in false positives", "figure": "29% more cancers detected; 44% less screen-reading workload", "detail": "The MASAI trial randomised over 100,000 women in Sweden to AI-supported screening or standard double reading. The AI arm detected 29% more cancers without more false positives, and radiologists' screen-reading workload fell by 44%.", "as_of": "2026-01-29", "checked": "2026-09-06", "source": { "title": "AI-supported mammography screening results in fewer aggressive and advanced breast cancers, finds full results from first randomized controlled trial (The Lancet)", "url": "https://www.eurekalert.org/news-releases/1114399", "publisher": "The Lancet, via EurekAlert" }, "also": [] }, { "id": "masai-interval", "area": "Medicine", "headline": "The same trial found 12% fewer cancers surfacing between screenings", "figure": "1.55 vs 1.76 interval cancers per 1,000 women", "detail": "Interval cancers, the ones found between screening rounds, ran at 1.55 per 1,000 women in the AI group against 1.76 per 1,000 in the control group, a 12% reduction, with fewer of them having unfavourable characteristics.", "as_of": "2026-01-29", "checked": "2026-09-06", "source": { "title": "Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading (MASAI)", "url": "https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(25)02464-X/abstract", "publisher": "The Lancet" }, "also": [] }, { "id": "rentosertib", "area": "Medicine", "headline": "The first AI-discovered drug to report a randomised Phase 2a result improved lung function where placebo declined", "figure": "FVC +98.4 mL (60 mg daily) vs -20.3 mL (placebo) at 12 weeks", "detail": "Rentosertib, a TNIK inhibitor whose target and molecule were produced with Insilico Medicine's generative AI platform, was tested in 71 patients with idiopathic pulmonary fibrosis across 22 sites. At 12 weeks the 60 mg once-daily arm gained a mean 98.4 mL of forced vital capacity while placebo lost 20.3 mL. Published in Nature Medicine.", "as_of": "2025-06-03", "checked": "2026-09-06", "source": { "title": "A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial (Nature Medicine)", "url": "https://www.eurekalert.org/news-releases/1086096", "publisher": "Nature Medicine, via EurekAlert" }, "also": [] }, { "id": "fda-devices", "area": "Medicine", "headline": "More than 1,600 AI-enabled medical devices are on the FDA's authorised list", "figure": "1,600+ entries; most recent decision dated 2026-09-04", "detail": "The U.S. Food and Drug Administration keeps a public list of AI-enabled medical devices it has authorised for marketing. Counted on 2026-09-06, the list held over 1,600 entries, with decisions running from 1995 to 4 September 2026.", "as_of": "2026-09-04", "checked": "2026-09-06", "source": { "title": "Artificial Intelligence-Enabled Medical Devices (FDA list)", "url": "https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices", "publisher": "U.S. Food and Drug Administration" }, "also": [] }, { "id": "alphafold-db", "area": "Science", "headline": "The AlphaFold database holds over 214 million predicted protein structures, used by more than 4.5 million people", "figure": "214,000,000+ structures; 4,501,953 total users", "detail": "The AlphaFold Protein Structure Database, built by EMBL-EBI and Google DeepMind, archived over 214 million predicted structures as of September 2023 and, per its 2025 update paper, had been used by 4,501,953 people. It is free and open.", "as_of": "2025-11-01", "checked": "2026-09-06", "source": { "title": "AlphaFold Protein Structure Database 2025: a redesigned interface and updated structural coverage (Nucleic Acids Research)", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12807749/", "publisher": "Nucleic Acids Research" }, "also": [ { "title": "AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10767828/" } ] }, { "id": "nobel-2024", "area": "Science", "headline": "The 2024 Nobel Prizes in Chemistry and Physics went to AI work", "figure": "Chemistry: protein structure prediction and design; Physics: neural networks", "detail": "The 2024 Nobel Prize in Chemistry was awarded to Demis Hassabis and John Jumper \"for protein structure prediction\" and David Baker \"for computational protein design\". The 2024 Nobel Prize in Physics went to John J. Hopfield and Geoffrey Hinton \"for foundational discoveries and inventions that enable machine learning with artificial neural networks\".", "as_of": "2024-10-09", "checked": "2026-09-06", "source": { "title": "All Nobel Prizes 2024", "url": "https://www.nobelprize.org/all-nobel-prizes-2024/", "publisher": "The Nobel Foundation" }, "also": [] }, { "id": "gnome", "area": "Science", "headline": "One deep-learning paper predicted 2.2 million new crystals, 380,000 of them stable", "figure": "2.2 million new crystals; 380,000 stable; known stable materials rose from ~48,000 to 421,000", "detail": "Google DeepMind's GNoME model, published in Nature, predicted 2.2 million new crystal structures, of which 380,000 are the most stable and candidates for synthesis. DeepMind describes it as nearly 800 years' worth of knowledge at the previous pace of discovery.", "as_of": "2023-11-29", "checked": "2026-09-06", "source": { "title": "Millions of new materials discovered with deep learning", "url": "https://deepmind.google/blog/millions-of-new-materials-discovered-with-deep-learning/", "publisher": "Google DeepMind" }, "also": [ { "title": "Scaling deep learning for materials discovery (Nature)", "url": "https://www.nature.com/articles/s41586-023-06735-9" } ] }, { "id": "imo-gold", "area": "Science", "headline": "An AI system reached gold-medal standard at the 2025 International Mathematical Olympiad, graded by the IMO itself", "figure": "5 of 6 problems, 35 of 42 points, inside the 4.5-hour limit", "detail": "An advanced version of Gemini with Deep Think solved five of the six IMO 2025 problems, scoring 35 points, the gold-medal threshold. The solutions were officially graded and certified by IMO coordinators using the same criteria as for students, and produced in natural language within the 4.5-hour window.", "as_of": "2025-07-21", "checked": "2026-09-06", "source": { "title": "Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad", "url": "https://deepmind.google/blog/advanced-version-of-gemini-with-deep-think-officially-achieves-gold-medal-standard-at-the-international-mathematical-olympiad/", "publisher": "Google DeepMind" }, "also": [] }, { "id": "gencast", "area": "Science", "headline": "An AI weather model beat the world's leading operational ensemble forecast on 97.2% of tested targets", "figure": "97.2% of targets, forecasts up to 15 days ahead", "detail": "GenCast, published in Nature, produced probabilistic forecasts that were more accurate than ECMWF's ENS system on 97.2% of the targets tested, out to 15 days, including extreme weather.", "as_of": "2024-12-04", "checked": "2026-09-06", "source": { "title": "GenCast predicts weather and the risks of extreme conditions with state-of-the-art accuracy", "url": "https://deepmind.google/blog/gencast-predicts-weather-and-the-risks-of-extreme-conditions-with-sota-accuracy/", "publisher": "Google DeepMind" }, "also": [ { "title": "Probabilistic weather forecasting with machine learning (Nature)", "url": "https://www.nature.com/articles/s41586-024-08252-9" } ] }, { "id": "aifs-operational", "area": "Disasters and weather", "headline": "Europe's weather centre put an AI forecast model into operations, using about 1,000 times less energy per forecast", "figure": "Operational 25 February 2025; gains of up to 20%; ~1,000x less energy per forecast", "detail": "ECMWF took its Artificial Intelligence Forecasting System (AIFS) into operations on 25 February 2025, running alongside its physics-based model. ECMWF reports gains of up to 20% on measures such as tropical cyclone tracks and roughly 1,000 times less energy per forecast.", "as_of": "2025-02-25", "checked": "2026-09-06", "source": { "title": "ECMWF's AI forecasts become operational", "url": "https://www.ecmwf.int/en/about/media-centre/news/2025/ecmwfs-ai-forecasts-become-operational", "publisher": "ECMWF" }, "also": [] }, { "id": "flood-hub", "area": "Disasters and weather", "headline": "AI river-flood forecasts cover 100 countries where 700 million people live, seven days ahead", "figure": "100 countries; 700 million people; 7-day lead time", "detail": "Google expanded its AI riverine flood forecasting to 100 countries covering areas where 700 million people live, with the model reaching the same accuracy at seven days as the previous model had at five. The underlying method was published in Nature in 2024.", "as_of": "2024-11-11", "checked": "2026-09-06", "source": { "title": "How Google helps others with AI flood forecasting", "url": "https://blog.google/innovation-and-ai/products/expanding-flood-forecasting-coverage-helping-partners/", "publisher": "Google" }, "also": [ { "title": "Global prediction of extreme floods in ungauged watersheds (Nature)", "url": "https://www.nature.com/articles/s41586-024-07145-1" } ] }, { "id": "alertcalifornia", "area": "Disasters and weather", "headline": "California's AI camera network detected 636 wildfires in 2024 before anyone called 911", "figure": "1,668 fires picked up by cameras; 636 (38%) before any 911 call; 1,211 cameras", "detail": "CAL FIRE's deputy chief for fire intelligence, Brian York, said that of 7,553 wildfires in CAL FIRE's jurisdiction in 2024, 1,668 were picked up by the ALERTCalifornia AI cameras and 636, or 38%, were detected before any person called 911. The network runs 1,211 cameras on peaks across the state.", "as_of": "2025-09-24", "checked": "2026-09-06", "source": { "title": "Over 1,000 AI cameras help spot Calif. wildfires before the first 911 call", "url": "https://www.gov1.com/emergency-management/over-1-000-ai-cameras-help-spot-calif-wildfires-before-the-first-911-call", "publisher": "Gov1 (quoting CAL FIRE)" }, "also": [] }, { "id": "be-my-eyes", "area": "Access and language", "headline": "More than 1 million blind or low-vision people use Be My Eyes, with AI answering alongside 10 million volunteers", "figure": "1,000,000+ users; 10,000,000+ volunteers; 150+ countries; 180+ languages", "detail": "On 12 March 2026 Be My Eyes announced more than 1 million blind or low-vision users and over 10 million volunteers, across more than 150 countries and 180 languages. Its Be My AI tool describes what the camera sees so users are not dependent on a volunteer being free.", "as_of": "2026-03-12", "checked": "2026-09-06", "source": { "title": "Be My Eyes Reaches 1 Million Blind and Low-Vision Users and 10 Million Volunteers", "url": "https://www.bemyeyes.com/news/be-my-eyes-reaches-1-million-blind-and-low-vision-users-and-10-million-volunteers/", "publisher": "Be My Eyes" }, "also": [] }, { "id": "translate-110", "area": "Access and language", "headline": "A single AI update added 110 languages to Google Translate, spoken by 614 million people", "figure": "110 languages; 614 million speakers; ~8% of the world's population", "detail": "Using the PaLM 2 language model, Google added 110 languages to Translate in its largest expansion ever, representing more than 614 million speakers, around 8% of the world's population, including languages such as Cantonese, Tamazight and Wolof.", "as_of": "2024-06-27", "checked": "2026-09-06", "source": { "title": "110 new languages are coming to Google Translate", "url": "https://blog.google/products-and-platforms/products/translate/google-translate-new-languages-2024/", "publisher": "Google" }, "also": [] }, { "id": "nllb-200", "area": "Access and language", "headline": "One open model translates 200 languages, 55 of them African, at 44% higher quality than the previous best", "figure": "200 languages; +44% average BLEU over prior state of the art; 55 African languages", "detail": "Meta's No Language Left Behind model, NLLB-200, translates between 200 languages with BLEU scores that improve on the previous state of the art by an average of 44%, and supports 55 African languages with high-quality results where widely used tools covered fewer than 25.", "as_of": "2022-07-06", "checked": "2026-09-06", "source": { "title": "200 languages within a single AI model: A breakthrough in high-quality machine translation", "url": "https://ai.meta.com/blog/nllb-200-high-quality-machine-translation/", "publisher": "Meta AI" }, "also": [] }, { "id": "pew-2026", "area": "Everyday use", "headline": "Half of U.S. adults now use AI chatbots, and about a quarter use them daily", "figure": "49% use AI chatbots (33% in 2024); ~24% daily; ChatGPT 44%", "detail": "Pew Research Center surveyed 5,119 U.S. adults from 17 to 23 February 2026: 49% use AI chatbots, up from 33% in 2024; about a quarter use them daily, including 12% several times a day. The most common uses were searching for information (42%) and work tasks (38% of employed adults).", "as_of": "2026-06-17", "checked": "2026-09-06", "source": { "title": "Americans and AI 2026: Chatbots, Smart Devices and Views on Impact", "url": "https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/", "publisher": "Pew Research Center" }, "also": [] } ] }