{ "dataset": "Where AI training data comes from", "description": "Quantified facts about the sources of AI training data, by kind of data: who holds it, how much exists and the channel it reaches buyers through.", "reviewed": "2026-10-16", "license": { "name": "CC BY 4.0", "url": "https://creativecommons.org/licenses/by/4.0/", "covers": "our compilation of the tables", "excludes": "quotes keep their sources' terms, and article text is not covered" }, "rows": [ { "id": "bls-customer-service-reps-2025", "dataType": "audio-speech", "source": "US contact centres (enterprises and BPOs)", "fact": "The US had 2,666,000 customer service representative jobs in 2025, and BLS projects a 5% decline for 2025 to 2035.", "figure": "2,666,000", "channel": "Mostly not reaching buyers; recorded calls held by enterprises and BPOs", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.bls.gov/ooh/office-and-administrative-support/customer-service-representatives.htm", "sourceLabel": "US Bureau of Labor Statistics, Occupational Outlook Handbook: Customer Service Representatives", "quote": "Number of Jobs, 2025 2,666,000", "addedOn": "2026-10-11" }, { "id": "common-voice-v27-hours-languages", "dataType": "audio-speech", "source": "Mozilla Common Voice (volunteer speech)", "fact": "Common Voice Scripted Speech v27.0 (data cut-off 2026-09) has 42,593 total hours, 29,295 validated hours and 295 languages.", "figure": "42593 total hours; 29295 validated hours; 295 languages", "channel": "Open download (CC0)", "date": "2026-09", "readOn": "2026-10-16", "basis": "stated", "url": "https://github.com/common-voice/cv-dataset", "sourceLabel": "Mozilla Common Voice, cv-dataset README", "quote": "Scripted Speech SCS Active 27 v27.0 295 ... Scripted Speech: Total & Validated Hours ... 41792,42388,42593] ... 28893,29295]", "addedOn": "2026-10-11" }, { "id": "five9-call-minutes-recorded", "dataType": "audio-speech", "source": "Contact centres on Five9 (4,000+ customers)", "fact": "Five9, one contact-centre platform, says 14+ billion call minutes are recorded on it each year for its 4,000+ customers.", "figure": "14+ billion call minutes recorded annually", "channel": "Not reaching AI buyers: recordings belong to the call centres' customers and are held under processor terms", "date": "2026-10-11", "readOn": "2026-10-16", "basis": "stated", "url": "https://www.five9.com/company/about", "sourceLabel": "Five9, About", "quote": "4,000+ enterprise, mid-market, and SMB customers worldwide 14+ billion call minutes recorded annually", "addedOn": "2026-10-11" }, { "id": "listennotes-podcasts-episodes-2026", "dataType": "audio-speech", "source": "Podcasters worldwide", "fact": "Listen Notes counts at least 3,824,033 podcasts and 195,710,248 episodes.", "figure": "3,824,033 podcasts; 195,710,248 episodes", "channel": "Not yet reaching buyers at scale; via brokers and podcast networks", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "listing", "url": "https://www.listennotes.com/podcast-stats/", "sourceLabel": "Listen Notes, Podcast Stats", "quote": "There are at least 3,824,033 podcasts* and 195,710,248 episodes in the world.", "addedOn": "2026-10-11" }, { "id": "shaip-english-speech-catalogue-hours", "dataType": "audio-speech", "source": "Shaip (speech data vendor)", "fact": "Shaip's speech catalogue lists 217,795 hours of English speech across 61 datasets, and 55 call-centre datasets.", "figure": "217,795 hours", "channel": "Vendor off-the-shelf licence", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://www.shaip.com/offerings/speech-data-catalog/", "sourceLabel": "Shaip, Speech Data Catalog", "quote": "217,795 hours · 61 sets", "addedOn": "2026-10-11" }, { "id": "spotify-7m-podcasts-700k-audiobooks", "dataType": "audio-speech", "source": "Spotify (podcast and audiobook catalogue)", "fact": "Spotify hosts 7 million podcast titles and 700,000 audiobooks.", "figure": "7 million podcast titles; 700,000 audiobooks", "channel": "Platform-held; rights sit with publishers and creators", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://newsroom.spotify.com/company-info/", "sourceLabel": "Spotify Newsroom, About Spotify", "quote": "over 100 million tracks, 7 million podcast titles, and 700,000 audiobooks on Spotify", "addedOn": "2026-10-11" }, { "id": "de-destatis-2024-legal-units-50-plus-employees", "dataType": "company-records", "source": "German legal units with 250+ (and 50-249) dependent employees", "fact": "In reporting year 2024, Germany had 18,632 legal units with 250 or more dependent employees and 84,701 with 50 to under 250, out of 3,543,865 in total.", "figure": "18 632", "channel": "Mostly not yet reaching buyers", "date": "2025-12-01", "readOn": "2026-10-16", "basis": "stated", "url": "https://www.destatis.de/DE/Themen/Branchen-Unternehmen/Unternehmen/Unternehmensregister/Tabellen/unternehmen-beschaeftigtengroessenklassen-wz08.html", "sourceLabel": "Statistisches Bundesamt (Destatis), Rechtliche Einheiten und abhängig Beschäftigte nach Beschäftigtengrößenklassen und Wirtschaftsabschnitten", "quote": "abhängig Beschäftigten 0 bis 10 10 bis 50 50 bis 250 250 und mehr ... B bis N, P bis S Insgesamt 3 543 865 40 083 660 3 011 214 5 603 979 429 318 8 737 488 84 701 8 452 481 18 632 17 289 712", "addedOn": "2026-10-11" }, { "id": "eu-sbs-2023-medium-and-large-enterprises", "dataType": "company-records", "source": "EU enterprises with 50-249 and 250+ persons employed", "fact": "In 2023, the EU business economy had 53,000 large enterprises (more than 249 persons employed) and 246,000 medium-sized enterprises (50-249), per preliminary SBS data.", "figure": "53 000", "channel": "Mostly not yet reaching buyers", "date": "2024-12-05", "readOn": "2026-10-11", "basis": "stated", "url": "https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20241205-1", "sourceLabel": "Eurostat, Large businesses make up only 0.2% of EU enterprises", "quote": "Large enterprises (more than 249 persons employed) represented only 0.2% of the total number of enterprises in the EU’s business economy (53 000 enterprises). ... Medium-sized enterprises (50-249 persons employed) constituted a small share of the total number of enterprises (0.8%, 246 000 enterprises)", "addedOn": "2026-10-11" }, { "id": "micro1-1b-enterprise-data-acquisition", "dataType": "company-records", "source": "micro1 (buyer)", "fact": "micro1 committed to spend $1B over the next 12 months to buy and license enterprise operational data through its Company Data Partnerships program, financed by Citi and Hercules Capital.", "figure": "$1B", "channel": "Direct deal", "date": "2026-10-09", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.micro1.ai/blog/micro1-commits-1b-to-enterprise-data-acquisition", "sourceLabel": "micro1 blog, micro1 commits $1B to enterprise data acquisition", "quote": "micro1 announces commitment to spend $1B on enterprise data acquisition over the next 12 months", "addedOn": "2026-10-11" }, { "id": "simpleclosure-workspace-tool-shares-sep-2026", "dataType": "company-records", "source": "Startup workspaces evaluated for sale (Slack, Drive, GitHub)", "fact": "Of the startup workspaces SimpleClosure evaluated for sale, 95% use Slack, 87% Google Drive and 82% GitHub.", "figure": "95%", "channel": "Wind-down service marketplace", "date": "2026-09-03", "readOn": "2026-10-11", "basis": "vendor", "url": "https://simpleclosure.com/blog/insights/what-your-assets-are-worth/", "sourceLabel": "SimpleClosure, What Your Assets Are Worth", "quote": "Among the workspaces we evaluated, 95% use Slack, 87% Google Drive and 82% GitHub.", "addedOn": "2026-10-11" }, { "id": "uk-bpe-2025-medium-and-large-businesses", "dataType": "company-records", "source": "UK medium (50-249) and large (250+) private businesses", "fact": "At the start of 2025, the UK had 38,435 medium-sized private sector businesses (50 to 249 employees) and 8,335 large ones (250 or more).", "figure": "38,435", "channel": "Mostly not yet reaching buyers", "date": "2025-10-02", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.gov.uk/government/statistics/business-population-estimates-2025/business-population-estimates-for-the-uk-and-regions-2025-statistical-release", "sourceLabel": "Department for Business and Trade, Business population estimates for the UK and regions 2025: statistical release", "quote": "38,435 businesses were medium-sized (50 to 249 employees) 8,335 businesses were large (250 or more employees)", "addedOn": "2026-10-11" }, { "id": "us-epiq-commercial-bankruptcies-jan-sep-2026", "dataType": "company-records", "source": "US businesses filing for bankruptcy", "fact": "From January 1 to September 30, 2026, there were 25,467 US commercial bankruptcy filings, up 7% from 23,789 in the same period of 2025 (September 2026 alone: 2,627).", "figure": "25,467", "channel": "Wind-down or bankruptcy asset sale, mostly not reaching data buyers", "date": "2026-10-09", "readOn": "2026-10-11", "basis": "reported", "url": "https://www.abfjournal.com/epiq-commercial-chapter-11-filings-decrease-16-y-y-in-september/", "sourceLabel": "ABF Journal, Epiq: Commercial Chapter 11 Filings Decrease 16% Y/Y in September", "quote": "Key bankruptcy statistics from Jan. 1 to Sept. 30, 2026, provided by Epiq AACER include: 25,467 overall commercial filings, a 7% increase from 2025 (23,789)", "addedOn": "2026-10-11" }, { "id": "us-susb-2022-firms-100-499-employees", "dataType": "company-records", "source": "US firms with 100-499 employees", "fact": "In reference year 2022, the United States had 93,341 employer firms with 100 to 499 employees.", "figure": "93341", "channel": "Mostly not yet reaching buyers (a few through direct deals or brokers)", "date": "2025-04-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://www2.census.gov/programs-surveys/susb/tables/2022/us_state_6digitnaics_2022.xlsx", "sourceLabel": "U.S. Census Bureau, 2022 SUSB Annual Data Tables by Establishment Industry (us_state_6digitnaics_2022.xlsx)", "quote": "00 | United States | -- | Total | 07: 100-499 employees | 93341", "addedOn": "2026-10-11" }, { "id": "us-susb-2022-firms-20-99-employees", "dataType": "company-records", "source": "US firms with 20-99 employees", "fact": "In reference year 2022, the United States had 561,160 employer firms with 20 to 99 employees.", "figure": "561160", "channel": "Mostly not yet reaching buyers (a few through direct deals or brokers)", "date": "2025-04-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://www2.census.gov/programs-surveys/susb/tables/2022/us_state_6digitnaics_2022.xlsx", "sourceLabel": "U.S. Census Bureau, 2022 SUSB Annual Data Tables by Establishment Industry (us_state_6digitnaics_2022.xlsx)", "quote": "00 | United States | -- | Total | 06: 20-99 employees | 561160", "addedOn": "2026-10-11" }, { "id": "us-susb-2022-firms-500-plus-employees", "dataType": "company-records", "source": "US firms with 500+ employees", "fact": "In reference year 2022, the United States had 21,041 employer firms with 500 or more employees (2,262 of them with 5,000+ in the detailed-size table).", "figure": "21041", "channel": "Mostly not yet reaching buyers (a few through direct deals or brokers)", "date": "2025-04-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://www2.census.gov/programs-surveys/susb/tables/2022/us_state_6digitnaics_2022.xlsx", "sourceLabel": "U.S. Census Bureau, 2022 SUSB Annual Data Tables by Establishment Industry (us_state_6digitnaics_2022.xlsx)", "quote": "00 | United States | -- | Total | 09: 500+ employees | 21041", "addedOn": "2026-10-11" }, { "id": "aws-data-exchange-products-providers", "dataType": "datasets-environments", "source": "AWS Data Exchange", "fact": "AWS Data Exchange lists more than 3,500 data products from over 300 providers, and over 1,000 of them are free.", "figure": "more than 3,500 products from over 300 providers", "channel": "Cloud marketplace", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://aws.amazon.com/data-exchange/why-aws-data-exchange/", "sourceLabel": "AWS, Why AWS Data Exchange", "quote": "AWS Data Exchange is the only data marketplace with more than 3,500 products from over 300 providers delivered", "addedOn": "2026-10-11" }, { "id": "data-europa-eu-dataset-count", "dataType": "datasets-environments", "source": "Publications Office of the EU (data.europa.eu)", "fact": "The European data portal lists 1,751,247 datasets from 212 catalogues in 36 countries.", "figure": "1 751 247 datasets", "channel": "Open download", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "stated", "url": "https://data.europa.eu/en", "sourceLabel": "data.europa.eu, The portal in numbers", "quote": "1 751 247 datasets 212 catalogues 36 countries", "addedOn": "2026-10-11" }, { "id": "data-gov-dataset-count", "dataType": "datasets-environments", "source": "US General Services Administration (Data.gov)", "fact": "Data.gov lists 586,902 datasets.", "figure": "586,902", "channel": "Open download", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "stated", "url": "https://catalog.data.gov/dataset", "sourceLabel": "Data.gov, Datasets catalog", "quote": "586,902 datasets available on Data.gov.", "addedOn": "2026-10-11" }, { "id": "datarade-providers-count", "dataType": "datasets-environments", "source": "Datarade", "fact": "Datarade lists data products from more than 2,700 data providers across more than 560 data categories.", "figure": "2700+ trusted data providers", "channel": "Broker marketplace", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://datarade.ai/", "sourceLabel": "Datarade, The #1 Global Data Marketplace", "quote": "The easy way to find and compare data products from 2700+ trusted data providers worldwide, for free.", "addedOn": "2026-10-11" }, { "id": "defined-ai-marketplace-datasets", "dataType": "datasets-environments", "source": "Defined.ai", "fact": "Defined.ai's dataset marketplace lists 819 off-the-shelf datasets. Its homepage cites a contributor crowd of 1.6M+ across 150+ countries.", "figure": "819 datasets", "channel": "Marketplace plus contributor crowd", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://defined.ai/datasets", "sourceLabel": "Defined.ai, Datasets", "quote": "Showing 10 of 819 datasets", "addedOn": "2026-10-11" }, { "id": "huggingface-public-datasets-count", "dataType": "datasets-environments", "source": "Hugging Face", "fact": "The Hugging Face Hub lists 1,092,755 datasets, almost all free to download.", "figure": "1,092,755", "channel": "Open download (hub)", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "listing", "url": "https://huggingface.co/datasets", "sourceLabel": "Hugging Face, Datasets", "quote": "Datasets 1,092,755", "addedOn": "2026-10-11" }, { "id": "huggingface-rl-environment-datasets", "dataType": "datasets-environments", "source": "Hugging Face Hub (datasets tagged rl-environment)", "fact": "Only 148 of the Hub's roughly 1.09 million datasets carry the rl-environment tag, so finished training environments are almost absent from the free supply.", "figure": "148", "channel": "Open download (hub)", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "listing", "url": "https://huggingface.co/datasets?other=rl-environment", "sourceLabel": "Hugging Face, Datasets filtered by tag rl-environment", "quote": "Datasets 148", "addedOn": "2026-10-11" }, { "id": "snowflake-marketplace-listings-providers", "dataType": "datasets-environments", "source": "Snowflake Marketplace", "fact": "Snowflake Marketplace lists more than 3,400 live data, agent and SaaS offerings from over 820 providers.", "figure": "over 820 providers, offering more than 3,400 live", "channel": "Cloud marketplace (in-platform sharing)", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://www.snowflake.com/en/product/features/marketplace/", "sourceLabel": "Snowflake, Snowflake Marketplace", "quote": "Snowflake Marketplace connects you to over 820 providers, offering more than 3,400 live, AI-ready data, agents and integrated SaaS solutions.", "addedOn": "2026-10-11" }, { "id": "zenodo-records-and-datasets", "dataType": "datasets-environments", "source": "Zenodo (CERN open research repository, university researchers)", "fact": "Zenodo's public API reports 7,458,033 records, of which 725,627 are typed as datasets and 306,429 as software.", "figure": "7,458,033 records; 725,627 datasets", "channel": "Open download under each record's own licence; no paid channel", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://zenodo.org/api/records?size=1&q=resource_type.type:dataset", "sourceLabel": "Zenodo REST API, records search (dataset filter returns \"total\": 725627; unfiltered returns \"total\": 7458033)", "quote": "\"total\": 725627", "addedOn": "2026-10-11" }, { "id": "bls-physicians-surgeons-jobs-2025", "dataType": "expert-work", "source": "US Bureau of Labor Statistics", "fact": "There were 862,800 physician and surgeon jobs in the US in 2025, a credentialed pool expert-data platforms recruit from.", "figure": "862,800", "channel": "Not reaching buyers (labour pool)", "date": "2026-08-27", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm", "sourceLabel": "BLS Occupational Outlook Handbook, Physicians and Surgeons", "quote": "Number of Jobs, 2025 862,800", "addedOn": "2026-10-11" }, { "id": "handshake-ai-monthly-payouts", "dataType": "expert-work", "source": "Handshake AI", "fact": "Handshake says its AI business reached about a $1B run rate and pays about $60M a month to over 30,000 people.", "figure": "~$60M to over 30K individuals every month", "channel": "Contributor platform", "date": "2026-06-15", "readOn": "2026-10-11", "basis": "listing", "url": "https://careers.base10.vc/companies/joinhandshake-com/jobs/82815909-fellow-experience-specialist-contract", "sourceLabel": "Handshake job posting (Base10 job board), Fellow Experience Specialist (Contract)", "quote": "We've grown from $0 to ~$1B run rate and pay ~$60M to over 30K individuals every month.", "addedOn": "2026-10-11" }, { "id": "mercor-experts-nearly-five-million", "dataType": "expert-work", "source": "Mercor", "fact": "Mercor says it has nearly five million experts on its platform.", "figure": "nearly five million experts", "channel": "Contributor platform", "date": "2026-06-25", "readOn": "2026-10-11", "basis": "stated", "url": "https://mercor.com/blog/update-on-mercor-security-incident/", "sourceLabel": "Mercor, Update on Mercor security incident", "quote": "Of our nearly five million experts, only a very limited subset had sensitive information affected.", "addedOn": "2026-10-11" }, { "id": "outlier-scale-contributors-payouts", "dataType": "expert-work", "source": "Outlier (Scale AI)", "fact": "Outlier says it has 900K+ AI trainers (master's holders, PhDs and graduates) and has paid experts more than $500M.", "figure": "900K+", "channel": "Contributor platform", "date": "2026-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://outlier.ai/", "sourceLabel": "Outlier, homepage", "quote": "Join 900K+ MAs, PhDs, and college graduates working and earning as AI trainers. ... $500M+ Paid to experts", "addedOn": "2026-10-11" }, { "id": "upwork-ai-related-gsv-q2-2026", "dataType": "expert-work", "source": "Upwork", "fact": "In Q2 2026, Upwork's gross services volume from AI-related work grew more than 22% year over year, while total gross services volume fell 4% to $966.4 million.", "figure": "more than 22%", "channel": "Contributor platform (general freelance marketplace)", "date": "2026-08-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/1627475/000162747526000046/upwork2q26-pressrelease.htm", "sourceLabel": "SEC EDGAR, Upwork Q2 2026 press release (8-K Ex. 99.1)", "quote": "GSV from AI-related work increased more than 22% year-over-year in Q2 2026. ... GSV(1) was $966.4 million, decreased 4% year-over-year", "addedOn": "2026-10-11" }, { "id": "aha-us-hospital-count", "dataType": "images", "source": "US hospitals (imaging and clinical record holders)", "fact": "The American Hospital Association counts 6,100 hospitals in the United States, each a holder of imaging archives and clinical records.", "figure": "6,100 hospitals", "channel": "De-identified data platforms and brokers (Segmed, Truveta, Epic Cosmos in-house); most archives not reaching AI buyers", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.aha.org/statistics/fast-facts-us-hospitals", "sourceLabel": "American Hospital Association, Fast Facts on U.S. Hospitals, 2026 (data from the 2024 AHA Annual Survey)", "quote": "There are 6,100 hospitals in the United States.", "addedOn": "2026-10-11" }, { "id": "getty-662m-visual-assets-600k-contributors", "dataType": "images", "source": "Getty Images (Getty, iStock, Unsplash)", "fact": "Getty Images has more than 662 million visual assets on its sites and distributes the content of more than 600,000 contributors and more than 360 premium partners.", "figure": "over 662 million; over 600,000", "channel": "Direct licence deals (data access for ML training)", "date": "2026-06-30", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/0001898496/000162828026055246/gety-20260630.htm", "sourceLabel": "Getty Images Holdings, Form 10-Q for the quarter ended June 30, 2026", "quote": "with over 662 million visual assets available through its industry-leading sites ... distributes the content of over 600,000 contributors and more than 360 premium content partners.", "addedOn": "2026-10-11" }, { "id": "getty-photographic-archive-150m-images", "dataType": "images", "source": "Getty Images archive (historical photography)", "fact": "Getty Images says it maintains one of the largest privately owned photographic archives in the world, with over 150 million images, separate from its creative and editorial stock.", "figure": "over 150 million images", "channel": "Direct licence deals; Getty reports data access and licensing inside its Other revenue", "date": "2026-06-30", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/0001898496/000162828026055246/gety-20260630.htm", "sourceLabel": "Getty Images Holdings, Form 10-Q for the quarter ended June 30, 2026", "quote": "Getty Images also maintains one of the largest and best privately-owned photographic archives in the world, with over 150 million images across geographies, periods, and verticals.", "addedOn": "2026-10-11" }, { "id": "laion-5b-image-text-pairs", "dataType": "images", "source": "LAION (web-crawled open dataset)", "fact": "LAION-5B contains 5.85 billion CLIP-filtered image-text pairs: 2.3 billion in English and 2.2 billion in more than 100 other languages.", "figure": "5,85B", "channel": "Open download (free substitute)", "date": "2022-03-31", "readOn": "2026-10-11", "basis": "stated", "url": "https://laion.ai/blog/laion-5b/", "sourceLabel": "LAION, LAION-5B: An open large-scale dataset for training next generation image-text models", "quote": "consisting of 5,85B CLIP-filtered image-text pairs ... 2,3B contain English language, 2,2B samples from 100+ other languages", "addedOn": "2026-10-11" }, { "id": "nci-imaging-data-commons-v23-size", "dataType": "images", "source": "NCI Imaging Data Commons (public research archive)", "fact": "NCI's Imaging Data Commons release V23 holds 95.33 TB across 160,199 imaging studies from 79,889 cases.", "figure": "95.33 TB", "channel": "Open download", "date": "2025-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://learn.canceridc.dev/data/data-release-notes", "sourceLabel": "NCI Imaging Data Commons, Data release notes (V23)", "quote": "Studies: 160,199 (+606) ... Cases: 79,889 (+355) ... Disk size: 95.33 TB (+2.22 TB)", "addedOn": "2026-10-11" }, { "id": "segmed-imaging-exams-platform", "dataType": "images", "source": "Segmed (aggregator of hospital and clinic imaging)", "fact": "Segmed's imaging platform holds approximately 150 million de-identified imaging exams across CT, MRI, X-ray, ultrasound and mammography, sourced from healthcare sites and licensed to AI developers.", "figure": "approximately 150 million de-identified imaging exams", "channel": "Broker: de-identified licensing to AI developers, also through Verily and Bayer platforms", "date": "2026-02-25", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.newswise.com/articles/segmed-partners-with-verily-to-expand-access-to-real-world-imaging-data", "sourceLabel": "Segmed press release via Newswise, Segmed partners with Verily to expand access to real-world imaging data", "quote": "which includes approximately 150 million de-identified imaging exams spanning modalities such as CT, MRI", "addedOn": "2026-10-11" }, { "id": "shutterstock-data-distribution-services-revenue-q2-2026", "dataType": "images", "source": "Shutterstock (stock images, footage, music, 3D)", "fact": "Shutterstock's Data, Distribution, and Services revenue (licensing its image, footage and music library and metadata, including to AI developers) was $56.1 million in Q2 2026 against $67.2 million in Q2 2025, and $77.2 million in the first half of 2026 against $106.9 million.", "figure": "$56.1 million (56,137 in thousands)", "channel": "Direct licence deals (data and metadata to AI developers)", "date": "2026-06-30", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/1549346/000154934626000029/sstk-20260630.htm", "sourceLabel": "Shutterstock, Form 10-Q for the quarter ended June 30, 2026 (Note 10)", "quote": "Three Months Ended June 30, Six Months Ended June 30, 2026 2025 2026 2025 ... Data, Distribution, and Services 56,137 67,194 77,181 106,926", "addedOn": "2026-10-11" }, { "id": "us-ct-exams-2023-jama-im", "dataType": "images", "source": "US hospitals and imaging centres", "fact": "About 93 million CT examinations were performed on 61.5 million patients in the United States in 2023.", "figure": "Ninety-three million", "channel": "Mostly not reaching buyers; de-identified hospital data partnerships", "date": "2025-04-14", "readOn": "2026-10-11", "basis": "stated", "url": "https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2832778", "sourceLabel": "JAMA Internal Medicine, Projected Lifetime Cancer Risks From Current Computed Tomography Imaging", "quote": "Ninety-three million CT examinations were performed in 61 510 000 patients in the United States in 2023", "addedOn": "2026-10-11" }, { "id": "agibot-world-beta-trajectories", "dataType": "physical-world", "source": "AgiBot / OpenDriveLab", "fact": "AgiBot World Beta has 1,003,672 robot trajectories (about 43.8T of data) under a non-commercial CC BY-NC-SA 4.0 license.", "figure": "1,003,672 trajectories (~43.8T)", "channel": "Open download (non-commercial license)", "date": "2025-03", "readOn": "2026-10-11", "basis": "stated", "url": "https://github.com/OpenDriveLab/AgiBot-World", "sourceLabel": "GitHub, OpenDriveLab/AgiBot-World", "quote": "Our complete dataset featuring 1,003,672 trajectories (~43.8T) ... All the data and code within this repo are under CC BY-NC-SA 4.0.", "addedOn": "2026-10-11" }, { "id": "droid-dataset-hours", "dataType": "physical-world", "source": "DROID consortium (13 institutions)", "fact": "DROID holds 76,000 teleoperated demonstration trajectories, or 350 hours of robot interaction data.", "figure": "76k demonstration trajectories or 350h", "channel": "Open download", "date": "2024-03-19", "readOn": "2026-10-11", "basis": "stated", "url": "https://droid-dataset.github.io/", "sourceLabel": "DROID project site, A Large-Scale In-The-Wild Robot Manipulation Dataset", "quote": "a diverse robot manipulation dataset with 76k demonstration trajectories or 350h of interaction data", "addedOn": "2026-10-11" }, { "id": "ego4d-egocentric-hours", "dataType": "physical-world", "source": "Ego4D consortium (Meta and universities)", "fact": "Ego4D holds 3,670 hours of first-person daily-life video from 923 participants in 74 locations across 9 countries.", "figure": "3,670 hours", "channel": "Open download (licensed)", "date": "2021-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://ego4d-data.org/", "sourceLabel": "Ego4D, official dataset site", "quote": "3,670 hours of video collected by 923 unique participants from 74 worldwide locations in 9 different countries", "addedOn": "2026-10-11" }, { "id": "ifr-industrial-robot-installations-2025", "dataType": "physical-world", "source": "Factory owners worldwide (counted by IFR)", "fact": "Factories worldwide installed more than 600,000 new industrial robots in 2025.", "figure": "more than 600,000", "channel": "Not reaching buyers (factory data stays with operators)", "date": "2026-09-24", "readOn": "2026-10-11", "basis": "stated", "url": "https://ifr.org/ifr-press-releases/five-million-robots-now-operate-in-factories-globally", "sourceLabel": "International Federation of Robotics, Five Million Robots now Operate in Factories Globally", "quote": "Factories worldwide installed more than 600,000 new units over the year.", "addedOn": "2026-10-11" }, { "id": "ifr-industrial-robot-operational-stock-2025", "dataType": "physical-world", "source": "Factory owners worldwide (counted by IFR)", "fact": "About 5 million industrial robots were in operation worldwide at the end of 2025, up 9% on the year.", "figure": "5 million units", "channel": "Not reaching buyers (factory data stays with operators)", "date": "2026-09-24", "readOn": "2026-10-11", "basis": "stated", "url": "https://ifr.org/ifr-press-releases/five-million-robots-now-operate-in-factories-globally", "sourceLabel": "International Federation of Robotics, Five Million Robots now Operate in Factories Globally", "quote": "The global operational stock of industrial robots surged 9% to a record 5 million units in 2025.", "addedOn": "2026-10-11" }, { "id": "meta-vjepa2-1m-hours-video", "dataType": "physical-world", "source": "Meta (buyer)", "fact": "Meta pretrained V-JEPA 2 on over 1 million hours of internet video, then post-trained its robot variant on under 62 hours of DROID robot video.", "figure": "over 1 million hours", "channel": "In-house collection (buyer volume)", "date": "2025-06-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://arxiv.org/abs/2506.09985", "sourceLabel": "arXiv, V-JEPA 2 (Meta)", "quote": "on a video and image dataset comprising over 1 million hours of internet video ... using less than 62 hours of unlabeled robot videos from the Droid dataset", "addedOn": "2026-10-11" }, { "id": "nvidia-cosmos-pretraining-20m-hours", "dataType": "physical-world", "source": "NVIDIA (buyer)", "fact": "NVIDIA pretrained its Cosmos world models on 9,000 trillion tokens, including 20 million hours of driving, robotics, synthetic and related data.", "figure": "20 million hours", "channel": "In-house collection (buyer volume)", "date": "2025-01-09", "readOn": "2026-10-16", "basis": "stated", "url": "https://developer.nvidia.com/blog/advancing-physical-ai-with-nvidia-cosmos-world-foundation-model-platform", "sourceLabel": "NVIDIA Technical Blog, Advancing Physical AI with NVIDIA Cosmos World Foundation Model Platform", "quote": "trained on 9,000 trillion tokens–including 20 million hours of data", "addedOn": "2026-10-11" }, { "id": "open-x-embodiment-trajectories", "dataType": "physical-world", "source": "Open X-Embodiment collaboration (Google DeepMind and research labs)", "fact": "Open X-Embodiment pools more than 1 million real robot trajectories from 22 robot embodiments, collected by 21 institutions.", "figure": "1M+ real robot trajectories", "channel": "Open download", "date": "2023-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://robotics-transformer-x.github.io/", "sourceLabel": "Open X-Embodiment project site, Robotic Learning Datasets and RT-X Models", "quote": "It contains 1M+ real robot trajectories spanning 22 robot embodiments ... We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions", "addedOn": "2026-10-11" }, { "id": "tesla-fsd-supervised-10-billion-miles", "dataType": "physical-world", "source": "Tesla (customer fleet)", "fact": "Tesla said owners had driven over 10 billion miles on FSD (Supervised) as of early May 2026.", "figure": "Over 10 billion miles", "channel": "In-house collection", "date": "2026-05-03", "readOn": "2026-10-11", "basis": "stated", "url": "https://x.com/Tesla/status/2050991662076477504", "sourceLabel": "Tesla on X, Over 10 billion miles driven on FSD Supervised", "quote": "Over 10 billion miles driven on FSD Supervised", "addedOn": "2026-10-11" }, { "id": "waymo-rider-only-miles-june-2026", "dataType": "physical-world", "source": "Waymo", "fact": "Waymo had driven 271.3 million rider-only (no human driver) miles through June 2026.", "figure": "271.3M", "channel": "In-house collection", "date": "2026-06", "readOn": "2026-10-11", "basis": "stated", "url": "https://waymo.com/safety/impact/", "sourceLabel": "Waymo, Safety Impact", "quote": "Through June 2026, Waymo has driven 271.3M rider-only miles without a human driver", "addedOn": "2026-10-11" }, { "id": "app-stores-app-count-oct-2026", "dataType": "source-code", "source": "App publishers on Google Play and the Apple App Store", "fact": "As of 16 October 2026, there are 2,635,872 apps on Google Play and 2,718,308 on the Apple App Store, each backed by a private codebase.", "figure": "2,635,872", "channel": "Direct deal (Google's confidential Play developer code pilot reported June 2026); otherwise not reaching buyers", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "vendor", "url": "https://42matters.com/stats", "sourceLabel": "42matters, Google Play vs iOS App Store | Store Stats for Mobile Apps", "quote": "As of 16 Oct 2026, there are 2,635,872 apps on Google Play and 2,718,308 apps on the Apple App Store", "addedOn": "2026-10-11" }, { "id": "bigcode-the-stack-v2-size", "dataType": "source-code", "source": "Permissively licensed public code (The Stack v2, built from Software Heritage)", "fact": "The Stack v2 contains over 3 billion files in 600+ languages, totalling 67.53TB uncompressed (about 900B training tokens per the card's table).", "figure": "67.53TB", "channel": "Open download (gated, terms of use)", "date": "2024-02", "readOn": "2026-10-11", "basis": "stated", "url": "https://huggingface.co/datasets/bigcode/the-stack-v2", "sourceLabel": "BigCode on Hugging Face, bigcode/the-stack-v2 dataset card", "quote": "The Stack v2 contains over 3B files in 600+ programming and markup languages. ... The total uncompressed size of all files is 67.53TB.", "addedOn": "2026-10-11" }, { "id": "cobol-lines-in-daily-use-2022", "dataType": "source-code", "source": "COBOL code in production at enterprises", "fact": "A Micro Focus global survey estimated 775 to 850 billion lines of COBOL in daily use.", "figure": "775-850 billion lines", "channel": "Not yet reaching buyers", "date": "2022-02-11", "readOn": "2026-10-11", "basis": "vendor", "url": "https://idm.net.au/node/13777", "sourceLabel": "IDM, COBOL Market 3X larger than Previously Estimated: Survey", "quote": "the amount of COBOL code in daily use increased significantly to 775-850 billion lines", "addedOn": "2026-10-11" }, { "id": "github-octoverse-2025-private-contribution-share", "dataType": "source-code", "source": "Private repositories on GitHub", "fact": "In 2025, 81.5% of contributions on GitHub (4.97B) happened in private repositories, though 63% of all repositories were public.", "figure": "81.5%", "channel": "Not yet reaching buyers (private, owner-controlled)", "date": "2025-10-28", "readOn": "2026-10-11", "basis": "stated", "url": "https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/", "sourceLabel": "GitHub Blog, Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1", "quote": "In 2025, 81.5% of contributions happened in private repositories, while 63% of all repositories were public.", "addedOn": "2026-10-11" }, { "id": "github-octoverse-2025-repositories", "dataType": "source-code", "source": "Repositories hosted on GitHub", "fact": "GitHub hosts 630 million repositories. 121 million were added in Octoverse year 2025 (Sep 2024 to Aug 2025), including 58 million new private repositories (up 33%).", "figure": "630M", "channel": "Public: open download or scrape; private: not yet reaching buyers", "date": "2025-10-28", "readOn": "2026-10-11", "basis": "stated", "url": "https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/", "sourceLabel": "GitHub Blog, Octoverse: A new developer joins GitHub every second as AI leads TypeScript to #1", "quote": "630M total repositories with +121M new repositories in 2025 marking our biggest year yet +58M private repositories (up 33%) underscore the activity happening outside of the public eye.", "addedOn": "2026-10-11" }, { "id": "simpleclosure-assethub-code-volume-sep-2026", "dataType": "source-code", "source": "Startups selling assets through SimpleClosure AssetHub", "fact": "The companies that offered assets on AssetHub since its April launch represent nearly 500 million lines of code across 3,000 repositories, with 1.7 million commits and 200,000 pull requests.", "figure": "nearly 500 million lines of code", "channel": "Wind-down service marketplace", "date": "2026-09-03", "readOn": "2026-10-11", "basis": "vendor", "url": "https://simpleclosure.com/blog/insights/what-your-assets-are-worth/", "sourceLabel": "SimpleClosure, What Your Assets Are Worth", "quote": "Since launching AssetHub in April, 350+ companies have offered assets through the marketplace, representing nearly 500 million lines of code across 3,000 repositories, 1.7 million commits, 200,000 pull requests, and 3,200 contributors.", "addedOn": "2026-10-11" }, { "id": "software-heritage-2025-unique-source-files", "dataType": "source-code", "source": "Public source code archived by Software Heritage", "fact": "The Software Heritage archive holds over 27 billion unique source files from 421 million projects.", "figure": "over 27 billion", "channel": "Open download (public archive; feeds datasets such as The Stack)", "date": "2026-01-16", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.softwareheritage.org/2026/01/16/software-heritage-activity-report-2025/", "sourceLabel": "Software Heritage, Software Heritage Activity Report: 2025", "quote": "The Archive now protects over 27 billion unique source files from 421 million projects.", "addedOn": "2026-10-11" }, { "id": "arxiv-total-submissions-oct-2026", "dataType": "text-archives", "source": "arXiv (Cornell, research preprints)", "fact": "arXiv has received 3,215,245 submissions since August 1991.", "figure": "3,215,245 submissions", "channel": "Open download (bulk access); each paper keeps its author's licence", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "stated", "url": "https://arxiv.org/stats/monthly_submissions", "sourceLabel": "arXiv, Monthly submissions statistics", "quote": "Total number of submissions as of October 16, 2026 = 3,215,245", "addedOn": "2026-10-11" }, { "id": "bartz-anthropic-works-list-482460", "dataType": "text-archives", "source": "Bartz v. Anthropic (N.D. Cal. 3:24-cv-05417)", "fact": "The Bartz v. Anthropic settlement covers 482,460 works on its Works List. By April 16, 2026, claims had been filed for 440,490 of them, a 91.3% claims rate.", "figure": "440,490 of the 482,460 Works", "channel": "Litigation settlement", "date": "2026-04-16", "readOn": "2026-10-11", "basis": "docket", "url": "https://storage.courtlistener.com/recap/gov.uscourts.cand.434709/gov.uscourts.cand.434709.643.0.pdf", "sourceLabel": "CourtListener RECAP, Dkt. 643 Plaintiffs' Updated Claims Report", "quote": "At present, the Class has filed claims for 440,490 of the 482,460 Works on the Works List, reflecting a claims rate of 91.3%.", "addedOn": "2026-10-11" }, { "id": "bowker-us-isbn-titles-2025", "dataType": "text-archives", "source": "Bowker (via Publishers Weekly)", "fact": "More than four million books with ISBNs were published in the US in 2025. More than 3.5 million were self-published, and 642,242 came from traditional publishers.", "figure": "more than four million; more than 3.5 million (self-published); 642,242 (traditional)", "channel": "Direct licence from publishers and authors, or litigation", "date": "2026-03-17", "readOn": "2026-10-11", "basis": "reported", "url": "https://www.publishersweekly.com/pw/print/20260323/99943-book-output-topped-4-million-in-2025.html", "sourceLabel": "Publishers Weekly, Book Output Topped Four Million in 2025", "quote": "The total number of books published in the U.S. in 2025 with ISBN numbers jumped 32.5% over 2024, to more than four million books, according to statistics compiled by Bowker. The increase was led by self-published works, for which the number of print and e-books (including those without BISAC codes) soared 38.7% to more than 3.5 million from 2.5 million in 2024. The number of traditionally published books released in 2025 rose 6.6%, to 642,242, from 602,428 in 2024.", "addedOn": "2026-10-11" }, { "id": "cloudflare-radar-2025-crawl-to-refer-anthropic-openai", "dataType": "text-archives", "source": "Cloudflare Radar", "fact": "In 2025, Anthropic's crawlers fetched up to 500,000 pages for every visitor they referred back, settling at about 25,000 to 100,000 per referral after May. OpenAI's ratio peaked at 3,700:1 in March.", "figure": "500,000:1; ~25,000:1 to ~100,000:1; 3,700:1", "channel": "Crawl, increasingly blocked or metered at the CDN", "date": "2025-12-15", "readOn": "2026-10-11", "basis": "stated", "url": "https://blog.cloudflare.com/radar-2025-year-in-review/", "sourceLabel": "Cloudflare Blog, The 2025 Cloudflare Radar Year in Review", "quote": "Anthropic had the highest crawl-to-refer ratios this year, reaching as much as 500,000:1, although they were quite erratic from January through May. ... After that, the ratios became more consistent, but remained higher than others, ranging from ~25,000:1 to ~100,000:1. OpenAI’s ratios over time were quite spiky, and reached as much as 3,700:1 in March.", "addedOn": "2026-10-11" }, { "id": "common-crawl-cc-main-2026-39-pages", "dataType": "text-archives", "source": "Common Crawl Foundation", "fact": "The September 2026 Common Crawl archive (CC-MAIN-2026-39) holds 2.17 billion web pages (361.4 TiB uncompressed) from 40.4 million hosts. Only 587.2 million of its URLs had never been crawled before.", "figure": "2.17 billion web pages (or 361.4 TiB of uncompressed content); 587.2 million new URLs", "channel": "Open download (free public bucket)", "date": "2026-09-19", "readOn": "2026-10-11", "basis": "stated", "url": "https://commoncrawl.org/blog/september-2026-crawl-archive-now-available", "sourceLabel": "Common Crawl, September 2026 Crawl Archive Now Available", "quote": "The data was crawled between September 4th and September 17th, and contains 2.17 billion web pages (or 361.4 TiB of uncompressed content). Page captures are from 40.4 million hosts or 33.2 million registered domains and include 587.2 million new URLs, not visited in any of our prior crawls.", "addedOn": "2026-10-11" }, { "id": "consent-in-crisis-c4-restricted-share", "dataType": "text-archives", "source": "Data Provenance Initiative", "fact": "Between 2023 and 2024, robots.txt restrictions cut off about 5% or more of all tokens in C4 and 28% or more of its most actively maintained sources, and 45% of C4 was restricted by terms of service.", "figure": "~5%+ of all tokens in C4, or 28%+ of the most actively maintained, critical sources; 45% of C4 (ToS)", "channel": "Crawl, now gated by robots.txt and terms of service", "date": "2024-07-20", "readOn": "2026-10-11", "basis": "stated", "url": "https://arxiv.org/abs/2407.14933", "sourceLabel": "arXiv, Consent in Crisis: The Rapid Decline of the AI Data Commons", "quote": "rendering ~5%+ of all tokens in C4, or 28%+ of the most actively maintained, critical sources in C4, fully restricted from use. For Terms of Service crawling restrictions, a full 45% of C4 is now restricted.", "addedOn": "2026-10-11" }, { "id": "crossref-registered-records-2026", "dataType": "text-archives", "source": "Crossref", "fact": "Crossref holds 189,130,320 registered metadata records, including 126,615,864 journal DOIs.", "figure": "189,130,320", "channel": "Open metadata API; full text by publisher licence", "date": "2026-10-16", "readOn": "2026-10-16", "basis": "stated", "url": "https://www.crossref.org/06members/53status.html", "sourceLabel": "Crossref, Crossref Stats Page", "quote": "Crossref Stats Page: Updated on Oct 16, 2026 MetaData Deposit Total Records in Database 189,130,320 ... Number of Journal DOIs 126,615,864", "addedOn": "2026-10-11" }, { "id": "epoch-public-human-text-stock-300t", "dataType": "text-archives", "source": "Epoch AI", "fact": "Epoch AI estimates the effective stock of public human-generated text at around 300 trillion tokens, and expects models to fully use it between 2026 and 2032 if current trends continue. This is still the June 2024 estimate: Epoch's page shows no 2025 or 2026 update.", "figure": "around 300 trillion tokens; between 2026 and 2032", "channel": "Open web crawl (the stock frontier models train on)", "date": "2024-06-06", "readOn": "2026-10-11", "basis": "stated", "url": "https://epoch.ai/blog/will-we-run-out-of-data-limits-of-llm-scaling-based-on-human-generated-data", "sourceLabel": "Epoch AI, Will we run out of data? Limits of LLM scaling based on human-generated data", "quote": "We estimate the effective stock of quality and repetition adjusted human-generated public text for AI training at around 300 trillion tokens. If trends continue, language models will fully utilize this stock between 2026 and 2032, or even earlier if intensely overtrained.", "addedOn": "2026-10-11" }, { "id": "institutional-books-1-harvard-242b-tokens", "dataType": "text-archives", "source": "Harvard Law School Library / Institutional Data Initiative", "fact": "Institutional Books 1.0 releases 983,004 public-domain volumes (242 billion tokens). They come from 1,075,899 Harvard Library volumes in over 250 languages that Google Books scanned.", "figure": "983,004 volumes, or 242B tokens", "channel": "Open download (public domain, Hugging Face)", "date": "2025-06-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://arxiv.org/abs/2506.08300", "sourceLabel": "arXiv, Institutional Books 1.0: A 242B token dataset from Harvard Library's collections", "quote": "originally spanning 1,075,899 volumes written in over 250 different languages for a total of approximately 250 billion tokens. As part of this initial release, the OCR-extracted text (original and post-processed) as well as the metadata (bibliographic, source, and generated) of the 983,004 volumes, or 242B tokens, identified as being in the public domain have been made available.", "addedOn": "2026-10-11" }, { "id": "internet-archive-wayback-one-trillion-pages", "dataType": "text-archives", "source": "Internet Archive", "fact": "The Wayback Machine passed one trillion preserved web pages in October 2025.", "figure": "one trillion web pages", "channel": "Public web archive (replay access, not a training licence)", "date": "2025-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://blog.archive.org/2025/10/31/one-trillion-web-pages-archived-internet-archive-celebrates-a-civilization-scale-milestone/", "sourceLabel": "Internet Archive Blogs, One Trillion Web Pages Archived", "quote": "Since 1996, the Wayback Machine has been saving the digital history of the internet. In October, it surpassed the threshold of preserving one trillion web pages", "addedOn": "2026-10-11" }, { "id": "newspapers-com-pages-held", "dataType": "text-archives", "source": "Newspapers.com (Ancestry)", "fact": "Newspapers.com, owned by Ancestry, holds 1,264,636,501 digitised newspaper pages from more than 31,500 newspapers dating back to 1690.", "figure": "1,264,636,501 newspaper pages", "channel": "Not reaching AI buyers: consumer subscription; no AI licence found and Ancestry's terms are reported to bar AI training", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.newspapers.com/", "sourceLabel": "Newspapers.com, homepage", "quote": "1,264,636,501 newspaper pages to explore.", "addedOn": "2026-10-11" }, { "id": "pmc-articles-available-fy25", "dataType": "text-archives", "source": "PubMed Central (US National Library of Medicine)", "fact": "PubMed Central had 11,327,616 full-text biomedical articles available in fiscal year 2025, up from 10,307,502 in FY24.", "figure": "11,327,616 articles", "channel": "Open download for the open-access subset; the rest under each publisher's licence", "date": "2025-09", "readOn": "2026-10-11", "basis": "stated", "url": "https://pmc.ncbi.nlm.nih.gov/about/intro/", "sourceLabel": "PubMed Central, PMC Overview", "quote": "Fiscal Year Articles Available Avg. Unique Sessions / Weekday FY25 11,327,616 3.8 million FY24 10,307,502", "addedOn": "2026-10-11" }, { "id": "prorata-partnered-publications", "dataType": "text-archives", "source": "ProRata.ai", "fact": "ProRata says more than 1,000 publications and content creators have partnered with it to license content to AI answers on a revenue-share basis.", "figure": "1000+ publications and content creators", "channel": "Revenue-share licensing platform", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "vendor", "url": "https://prorata.ai/", "sourceLabel": "ProRata.ai, homepage", "quote": "Join the 1000+ publications and content creators that have partnered with us.", "addedOn": "2026-10-11" }, { "id": "qwen3-pretraining-tokens", "dataType": "text-archives", "source": "Alibaba Qwen (buyer-side consumption)", "fact": "Alibaba's Qwen3 models were pretrained on approximately 36 trillion tokens covering 119 languages and dialects, nearly twice the amount used for Qwen2.5.", "figure": "approximately 36 trillion tokens", "channel": "Consumption benchmark: web crawl, licensed and synthetic data", "date": "2025-04-29", "readOn": "2026-10-11", "basis": "stated", "url": "https://qwenlm.github.io/blog/qwen3/", "sourceLabel": "Qwen team, Qwen3: Think Deeper, Act Faster", "quote": "Qwen3 uses nearly twice that amount, with approximately 36 trillion tokens covering 119 languages and dialects.", "addedOn": "2026-10-11" }, { "id": "reddit-other-revenue-fy2025", "dataType": "text-archives", "source": "Reddit, Inc.", "fact": "Reddit's other revenue, which is mostly content licensing, was $140.0 million in 2025, up from $114.7 million in 2024 and $15.2 million in 2023. In Q2 2026 it was $43.3 million.", "figure": "140,026 (in thousands)", "channel": "Direct licence deal", "date": "2026-02-06", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/1713445/000171344526000022/rddt-20251231.htm", "sourceLabel": "SEC EDGAR, Reddit Inc. Form 10-K FY2025", "quote": "Other revenue consists of revenue from content licensing and products sold directly to users. ... (in thousands) Advertising revenue $ 2,062,480 $ 1,185,456 $ 788,782 Other revenue 140,026 114,749 15,247", "addedOn": "2026-10-11" }, { "id": "rsl-standard-endorsers", "dataType": "text-archives", "source": "RSL Collective / RSL Technical Steering Committee", "fact": "More than 1,500 media organizations, brands and technology companies endorsed the RSL machine-readable AI licensing standard when version 1.0 launched.", "figure": "1500+ media organizations, brands, and technology companies", "channel": "Machine-readable licence standard and collective licensing", "date": "2025-12-10", "readOn": "2026-10-11", "basis": "stated", "url": "https://rslstandard.org/press/rsl-1-specification-2025", "sourceLabel": "RSL Standard, RSL AI Licensing 1.0 Now an Official Industry Standard", "quote": "1500+ media organizations, brands, and technology companies worldwide now endorse RSL; support spans billions of web pages, representing most high-quality internet content used in AI foundation model training", "addedOn": "2026-10-11" }, { "id": "stack-overflow-new-questions-sep-2026", "dataType": "text-archives", "source": "Stack Overflow (new public Q&A)", "fact": "Stack Overflow's own API counts 986 questions asked in September 2026 and still on the site, against 24,732 for September 2024 and 119,730 for September 2021 on the same query.", "figure": "986 questions in September 2026", "channel": "Direct API licence (OverflowAPI) to labs; the flow of new public text behind it is collapsing", "date": "2026-09", "readOn": "2026-10-11", "basis": "stated", "url": "https://api.stackexchange.com/2.3/questions?fromdate=1788220800&todate=1790812800&site=stackoverflow&filter=total", "sourceLabel": "Stack Exchange API 2.3, questions on stackoverflow created 1 to 30 September 2026, total filter", "quote": "{\"total\":986}", "addedOn": "2026-10-11" }, { "id": "stack-overflow-questions-total", "dataType": "text-archives", "source": "Stack Overflow (Prosus)", "fact": "Stack Overflow's public API reports 24,134,193 questions on the site in total.", "figure": "24,134,193 questions", "channel": "Direct API licence (OverflowAPI) to labs, plus open data dumps", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "listing", "url": "https://api.stackexchange.com/2.3/questions?site=stackoverflow&filter=total", "sourceLabel": "Stack Exchange API 2.3, questions on stackoverflow, total filter", "quote": "{\"total\":24134193}", "addedOn": "2026-10-11" }, { "id": "taus-translation-memory-sale-2024", "dataType": "text-archives", "source": "TAUS Data Association (pooled translation memories)", "fact": "TAUS put its pooled collection of about 7.4 billion words of non-public human translation data in 483 language pairs on sale for LLM training at discounts of more than 95% in March and April 2024.", "figure": "7.4 billion words, 483 language pairs", "channel": "Data marketplace sale (TAUS Data Marketplace)", "date": "2024-03-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://taus.net/resources/blog/taus-data-sale-to-boost-multilingual-llms", "sourceLabel": "TAUS, TAUS Data Sale to Boost Multilingual LLMs", "quote": "The 7.4 billion words on offer are all non-public, unique, human translation quality data covering 483 language pairs.", "addedOn": "2026-10-11" }, { "id": "liverpool-lecture-capture-hours", "dataType": "video", "source": "University of Liverpool (lecture recordings)", "fact": "One UK university alone has recorded and shared more than 85,000 hours of teaching through lecture capture since 2013.", "figure": "more than 85,000 hours", "channel": "Not reaching AI buyers: internal teaching platform (Panopto)", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://www.liverpool.ac.uk/it/lecture-capture/", "sourceLabel": "University of Liverpool IT Services, Lecture capture", "quote": "Stream Lecture Capture was introduced at the University in 2013 and has since been used to record and share more than 85,000 hours of teaching with our University Community.", "addedOn": "2026-10-11" }, { "id": "pond5-30-million-video-clips-100k-artists", "dataType": "video", "source": "Pond5 (stock footage marketplace, Shutterstock-owned)", "fact": "Pond5 passed 30 million licensable royalty-free video clips, with 100,000 artists uploading 15,000 new videos a day.", "figure": "30 million", "channel": "Contributor platform; reaches AI buyers through parent Shutterstock's data licensing", "date": "2022-01-19", "readOn": "2026-10-11", "basis": "reported", "url": "https://broadcastbeat.com/news/pond5-surpasses-a-record-30-million-licensable-videos", "sourceLabel": "Broadcast Beat, Pond5 Surpasses a Record 30 Million Licensable Videos", "quote": "has become the first stock media company to exceed 30 million licensable royalty-free video clips. ... With 100,000 artists around the globe uploading 15,000 new videos every day", "addedOn": "2026-10-11" }, { "id": "steam-games-released-2025-steamdb", "dataType": "video", "source": "Game studios publishing on Steam", "fact": "SteamDB counted 20,008 games released on Steam in 2025. Other trackers counted 20,282 to 20,558.", "figure": "20008", "channel": "Not yet reaching buyers at scale (studio by studio)", "date": "2026-01-08", "readOn": "2026-10-11", "basis": "reported", "url": "https://howtomarketagame.com/2026/01/08/how-many-games-were-released-in-2025/", "sourceLabel": "How To Market A Game, How many games were released in 2025", "quote": "20,558 – GameDiscoverCo ... 20,282 – VGInsights. ... 20008 – SteamDb", "addedOn": "2026-10-11" }, { "id": "vimeo-daily-uploads", "dataType": "video", "source": "Vimeo (business and creator video hosting)", "fact": "Vimeo says an average of more than 350,000 videos are uploaded to it every day, by 1.7 million or more paid subscribers in 190+ countries.", "figure": "350K+ videos a day", "channel": "Not reaching AI buyers: hosting platform; footage belongs to uploaders", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://vimeo.com/about", "sourceLabel": "Vimeo, About", "quote": "Daily video uploads to Vimeo Average number of videos uploaded to Vimeo every single day 350K+", "addedOn": "2026-10-11" }, { "id": "youtube-20-million-videos-uploaded-daily", "dataType": "video", "source": "YouTube (creators, platform)", "fact": "YouTube says an average of more than 20 million videos are uploaded to it every day.", "figure": "over 20 million", "channel": "Held by platform and creators; no general training licence", "date": "2026-10-11", "readOn": "2026-10-11", "basis": "stated", "url": "https://blog.youtube/press/", "sourceLabel": "YouTube Official Blog, YouTube for Press", "quote": "On average, there are over 20 million videos uploaded daily to YouTube.", "addedOn": "2026-10-11" }, { "readOn": "2026-10-14", "id": "troveo-paid-creators-50m-7m-hours", "dataType": "video", "source": "Video creators and media companies licensing through Troveo", "fact": "Troveo says it has paid more than $50 million to creators, counts about 7,000 content owners and has licensed more than 7 million hours of video. Its business-data page gives $20 million or more paid to licensors on the same day.", "figure": "$50M+ paid; 7K content owners; 7M+ hours licensed", "channel": "Licensed to AI labs through Troveo, a royalty on every use", "date": "2026-10-14", "basis": "stated", "url": "https://www.troveo.ai/content-owners", "sourceLabel": "Troveo, for content owners", "quote": "$ 50 M+ ... paid To creators globally 7 K ... of content owners Earning passive income 7 M+ ... hours Video content licensed", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "pangea-paid-code-owners-600k-380-codebases", "dataType": "source-code", "source": "Agencies and software companies licensing private code through Pangea.ai's partner", "fact": "Pangea says $600,000 has been paid to code owners across its network for 380 codebases, about $1,579 a codebase.", "figure": "$600,000; 380 codebases", "channel": "Non-exclusive code licences sold to AI labs by Pangea's unnamed partner", "date": "2026-10-14", "basis": "stated", "url": "https://referral-partner-guide-pangea.netlify.app/", "sourceLabel": "Pangea.ai, Code Licensing referral partner guide", "quote": "Pangea has already paid $600,000 to code owners across the network. ... would earn again on any of those 380 codebases that are licensed more than once.", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "labelbox-prism-100m-distributed-1k-companies", "dataType": "company-records", "source": "Companies licensing operational data through Labelbox Prism", "fact": "Labelbox Prism says it has distributed more than $100 million to over 1,000 companies licensing their data, with about 21 days to the first payment.", "figure": "$100M+; 1k+ companies; ~21 days to first wire", "channel": "Licensed through Labelbox Prism", "date": "2026-10-14", "basis": "stated", "url": "https://prism.labelbox.com/enterprise", "sourceLabel": "Labelbox Prism, enterprise data licensing", "quote": "$100M+ distributed 1k+ companies licensing ~21d to first wire", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "handshake-paid-out-to-partners-250m", "dataType": "company-records", "source": "Companies licensing their data through Handshake AI", "fact": "Handshake AI's company-data page says it has paid out $250 million to partners and has more than 1,000 partner companies; it quotes payouts of $100K to $4M a company. The page does not say whether the $250 million includes its expert workers.", "figure": "$250M paid out; 1K+ partner companies", "channel": "Company data licensed through Handshake AI", "date": "2026-10-14", "basis": "stated", "url": "https://joinhandshake.com/ai/data-partnerships/license", "sourceLabel": "Handshake AI, Get paid for your company data", "quote": "Payouts range from $100K to $4M. $250M Paid out to partners 1M+ Handshake employers 1K+ Partner companies", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "defined-ai-partner-revenue-600k-average", "dataType": "datasets-environments", "source": "Data owners selling datasets through Defined.ai", "fact": "Defined.ai says its data partners earn more than $600,000 on average, more than five have earned over $1 million, and its dataset sales grew 319% in a year.", "figure": "$600K+ average partner revenue; $1M+ for 5+ partners; 319% YoY", "channel": "Datasets sold through Defined.ai's marketplace and partnership programmes", "date": "2026-10-14", "basis": "stated", "url": "https://www.defined.ai/partnerships", "sourceLabel": "Defined.ai, Partnership Programs", "quote": "319% YoY growth in dataset sales $600K+ Average partner revenue $1M+ Revenue for 5+ partners", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "wirestock-paid-contributors-15m", "dataType": "images", "source": "Photographers and creators supplying Wirestock", "fact": "Wirestock, which supplies multimodal data to six of the largest foundation model makers, has paid $15 million to its contributors on an annual run-rate revenue of $40 million.", "figure": "$15 million paid out; $40 million run-rate revenue", "channel": "Contributor images and media sold to AI labs through Wirestock", "date": "2026-05-14", "basis": "reported", "url": "https://techcrunch.com/2026/05/14/wirestock-raises-23m-to-supply-multi-modal-data-to-ai-labs/", "sourceLabel": "TechCrunch, Wirestock raises $23M", "quote": "He noted the company currently has an annual run-rate revenue of $40 million and has so far paid out $15 million to its contributors.", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "figure-index-paid-contributors-35m", "dataType": "physical-world", "source": "People filming everyday tasks through Figure AI's Index app", "fact": "Figure AI's Index app shows $35 million paid to contributors, 28.5 million video uploads, 1 million app downloads and contributors in 111 countries. Figure said $15 million on 25 August 2026.", "figure": "$35M paid; 28.5M video uploads; 111 countries", "channel": "First-person task video captured and paid by the minute through Figure's Index app", "date": "2026-10-14", "basis": "stated", "url": "https://www.figure.ai/index-app", "sourceLabel": "Figure AI, Index app", "quote": "App Downloads 1M Video Uploads 28.5M Countries 111 Paid to Contributors $35M", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "scale-swe-bench-pro-18-startup-repositories", "dataType": "source-code", "source": "Early-stage startups' private repositories", "fact": "Scale AI's SWE-Bench Pro benchmark keeps a commercial set of 18 proprietary repositories from early-stage startups under formal partnership agreements; the paper says Scale purchased them.", "figure": "18 proprietary repositories", "channel": "Bought from startups by Scale AI for a commercial evaluation set", "date": "2025-09-21", "basis": "stated", "url": "https://arxiv.org/abs/2509.16941", "sourceLabel": "Scale AI, SWE-Bench Pro (arXiv 2509.16941)", "quote": "a commercial set of 18 proprietary repositories where we have formal partnership agreements with early-stage startups", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "wiley-ai-licensing-q1-fy2027", "dataType": "text-archives", "source": "Wiley's research and learning content", "fact": "Wiley's AI licensing revenue was $14 million in the quarter to 31 July 2026, against $29 million in the same quarter a year earlier.", "figure": "$14 million (quarter to July 2026); $29 million a year earlier", "channel": "Publisher content licensed to AI developers for training, commercial use and knowledge feeds", "date": "2026-09-03", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/107140/000162828026060252/ex991earningsrelease.htm", "sourceLabel": "John Wiley & Sons, Q1 fiscal 2027 earnings release (SEC)", "quote": "prior year AI licensing revenue of $29 million ... AI licensing revenue totaled $14 million for the quarter.", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "protege-transaction-volume-2026", "dataType": "datasets-environments", "source": "Data holders brokered by Protege (healthcare, media and enterprise records)", "fact": "Protege's CEO says its total transaction volume rose from $30 million last year to at least $100 million this year.", "figure": "$30 million (2025); at least $100 million (2026)", "channel": "Brokered by Protege to AI labs", "date": "2026-08-14", "basis": "reported", "url": "https://en.fnnews.com/news/202608141625145080", "sourceLabel": "Financial News (Korea), citing The Information", "quote": "Bobby Samuels, CEO of U.S. data brokerage startup Protege, said the company’s total transaction volume has jumped from $30 million last year to at least $100 million this year", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "simpleclosure-expects-10m-to-startups-2026", "dataType": "company-records", "source": "Closing startups listing code and workspace data on SimpleClosure AssetHub", "fact": "SimpleClosure expects to return more than $10 million to startups listing data on AssetHub by the end of 2026, after about $1 million across nearly 100 deals in its test year.", "figure": "more than $10 million (2026 forecast)", "channel": "Sold to AI buyers through SimpleClosure AssetHub", "date": "2026-09-03", "basis": "reported", "url": "https://www.upstartsmedia.com/p/exclusive-simpleclosure-data-what-ai-labs-buy", "sourceLabel": "Upstarts, SimpleClosure data on what AI labs buy", "quote": "the company expects to return more than $10 million to listing startups by the end of the year, SimpleClosure says.", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "google-content-offer-pilot-thousands-of-partners", "dataType": "datasets-environments", "source": "Owners of large specialised content collections (code repositories and others)", "fact": "Google says it has tested its Content Offer Pilot for over a year with thousands of partners worldwide, setting each fee in a one-to-one offer process; the deal is non-exclusive and the partner keeps its IP.", "figure": "thousands of partners; 90%+ would participate again", "channel": "Sold directly to Google through its Content Offer Pilot", "date": "2026-10-14", "basis": "stated", "url": "https://contentpilot.google.com/", "sourceLabel": "Google, Content Offer Pilot", "quote": "90%+ would participate again We’ve been testing this pilot for over a year with thousands of partners worldwide.", "addedOn": "2026-10-14" }, { "readOn": "2026-10-14", "id": "shutterstock-ai-licensing-revenue-2023", "dataType": "images", "source": "Shutterstock's image, video and music library and its contributors", "fact": "Shutterstock earned $104 million of AI licensing revenue in 2023 across all its AI customers.", "figure": "$104 million (2023)", "channel": "Library licensed to AI developers (OpenAI, Meta, Google, Amazon, Apple and others)", "date": "2023-12", "basis": "reported", "url": "https://llmpulse.ai/blog/ai-content-licensing-deals/", "sourceLabel": "LLM Pulse, AI content licensing deals 2023 to 2026", "quote": "Shutterstock: 104 million dollars of AI licensing revenue in 2023 alone, across all its AI customers.", "addedOn": "2026-10-14" }, { "id": "hilco-project-beacon-engineering-archive", "dataType": "company-records", "source": "An unnamed wireless-power technology company, offered through Hilco Global", "fact": "Hilco Global is selling, under NDA, a wireless-power company's engineering archive: about 1.25 million lines of company-authored code (6.0 million with vendor SDKs), 202 firmware projects and a 32.6 GB archive of 53,712 engineering emails from August 2017 to January 2024. No price, bid deadline or court case is stated; the listing does not mention AI buyers.", "figure": "53,712 engineering emails; 1.25MM+ lines of company-authored code", "channel": "Intangible-asset liquidator (Hilco Global) sale process under NDA", "date": "2026-09", "readOn": "2026-10-14", "basis": "listing", "url": "https://hilcoglobal.com/professional-services/intangible-assets/for-sale/project-beacon-wireless-powering-tech-code-technical-data-assets-opportunity/", "sourceLabel": "Hilco Global, Project Beacon: Wireless Powering Tech Code & Technical Data Assets Opportunity", "quote": "32.6 GB engineering email archive containing 53,712 messages from August 2017 through January 2024.", "addedOn": "2026-10-14" }, { "id": "spirit-deidentified-data-sale-scope-97m-consumers", "dataType": "company-records", "source": "Spirit Airlines estate (In re Spirit Aviation Holdings, Bankr. S.D.N.Y.)", "fact": "The Spirit enterprise data on sale to Google derives from about 190 million flights and touches the personal data of about 97 million passengers; after the 5 October 2026 narrowing the structured passenger systems are out and corporate email stays in, de-identified by Tonic.ai.", "figure": "about 190 million flights; 97 million consumers", "channel": "Chapter 11 section 363 sale (auction run by Davis Polk and PJT Partners), approval hearing 6 November 2026", "date": "2026-10-05", "readOn": "2026-10-14", "basis": "stated", "url": "https://storage.courtlistener.com/recap/gov.uscourts.nysb.331039/gov.uscourts.nysb.331039.1684.0.pdf", "sourceLabel": "In re Spirit Aviation Holdings, Dkt. 1684 (Consumer Privacy Ombudsman supplemental report)", "quote": "the Debtors proposed to sell enterprise data derived from the company’s operations that include “deidentified data” pertaining to approximately 190 million airline flights taken by Spirit passengers", "addedOn": "2026-10-14" }, { "id": "openai-purchased-hundreds-of-ui-gym-sites", "dataType": "datasets-environments", "source": "Website replicas (UI gyms) bought from environment vendors", "fact": "SemiAnalysis reports that UI gyms (website replicas for agent training) often cost about $20,000 per website and that OpenAI has bought hundreds of them for ChatGPT Agent.", "figure": "hundreds of sites", "channel": "Bespoke builds sold by environment vendors to labs", "date": "2026-01-06", "readOn": "2026-10-14", "basis": "reported", "url": "https://newsletter.semianalysis.com/p/rl-environments-and-rl-for-science", "sourceLabel": "SemiAnalysis, RL Environments and RL for Science: Data Foundries and Multi-Agent Architectures", "quote": "These 'UI gyms' often cost about $20,000 per website ... OpenAI has purchased hundreds of sites for ChatGPT Agent training and development.", "addedOn": "2026-10-14" }, { "id": "prime-intellect-400-environments-crowdsourced", "dataType": "datasets-environments", "source": "Open-source RL environments built through Prime Intellect bounties", "fact": "Prime Intellect crowdsourced more than 400 RL environments and evals through bounties in two months, with 80+ reviewed implementations.", "figure": "400+ environments", "channel": "Open release on Prime Intellect's Environments Hub", "date": "2025-10-27", "readOn": "2026-10-14", "basis": "stated", "url": "https://www.primeintellect.ai/blog/scaling-environments-program", "sourceLabel": "Prime Intellect, Scaling the Environments Program", "quote": "In the past 2 months, we've crowdsourced 400+ frontier RL environments & evals through bounties", "addedOn": "2026-10-14" }, { "id": "mifeng-1m-hours-embodiment-free-data", "dataType": "physical-world", "source": "Embodiment-free human manipulation data (UMI, egocentric) produced in China", "fact": "Chinese embodied-data service company Mifeng Technology (觅蜂科技) says it has produced over 1 million hours of embodiment-free data; Noematrix (穹彻智能) collected over 100,000 hours of UMI data and about 10,000 hours of egocentric data in about six months.", "figure": "1,000,000+ hours", "channel": "Sold to embodied-AI model developers by data service companies; collected in-house by model developers", "date": "2026-09-24", "readOn": "2026-10-14", "basis": "reported", "url": "https://www.cnfin.com/hg-lb/detail/20260924/4474923_1.html", "sourceLabel": "Xinhua Finance (cnfin.com), 从遥操作到无本体 具身智能数据迎有效性大考", "quote": "前不久，具身智能数据服务公司觅蜂科技宣布累计生产无本体数据超过100万小时。", "addedOn": "2026-10-14" }, { "id": "china-teleop-8-hours-for-1-valid-hour", "dataType": "physical-world", "source": "Real-robot teleoperation collection in China", "fact": "A year before September 2026, eight hours of robot work typically yielded one hour of valid data, per industry bodies quoted by Xinhua Finance.", "figure": "8 hours for 1 valid hour", "channel": "Data collection centres and in-house teleoperation", "date": "2026-09-24", "readOn": "2026-10-14", "basis": "reported", "url": "https://www.cnfin.com/hg-lb/detail/20260924/4474923_1.html", "sourceLabel": "Xinhua Finance (cnfin.com), 从遥操作到无本体 具身智能数据迎有效性大考", "quote": "一年前，工作人员带着机器人工作8个小时，往往只能产出1小时有效数据", "addedOn": "2026-10-14" }, { "id": "uthana-250-hours-studio-mocap-licensable", "dataType": "physical-world", "source": "Studio-grade human motion capture (Uthana)", "fact": "Uthana licenses 250+ hours of human motion captured with a 120-camera marker-based system in a AAA studio, for commercial model training under agreement.", "figure": "250+ hours", "channel": "Licensed directly by the vendor, price on request", "date": "2026-10-14", "readOn": "2026-10-14", "basis": "stated", "url": "https://uthana.com/solutions/data", "sourceLabel": "Uthana, Data", "quote": "License 250+ hours of human motion recorded with a 120-camera marker-based system in a AAA production studio.", "addedOn": "2026-10-14" }, { "id": "noitom-hiphi-617-hours-mocap", "dataType": "physical-world", "source": "High-precision motion capture for humanoids (Noitom HiPHI)", "fact": "Noitom released HiPHI, 617.5 hours of motion capture from 132 performers, free for non-commercial research with enterprise access through ModalityNet.", "figure": "617.5 hours", "channel": "Free for research on Hugging Face; commercial access by agreement", "date": "2026-08-20", "readOn": "2026-10-14", "basis": "reported", "url": "https://www.eweek.com/news/noitom-hiphi-motion-dataset-humanoid-ai-china-apac/", "sourceLabel": "eWeek, Noitom Releases 617-Hour HiPHI Motion Dataset", "quote": "Spanning 617.5 hours of high-precision motion capture, the benchmark, as Noitom described it,", "addedOn": "2026-10-14" }, { "id": "shutterstock-data-distribution-services-revenue-fy2025", "dataType": "images", "source": "Shutterstock (stock images, footage, music, 3D)", "fact": "Shutterstock's Data, Distribution, and Services revenue (which includes licences to the metadata of its images, footage, music and 3D models for machine learning and generative AI training) was $203.3 million in 2025, up 16% from $175.3 million in 2024 and $137.3 million in 2023.", "figure": "$203.3 million (203,264 in thousands)", "channel": "Direct licence deals (data and metadata to AI developers)", "date": "2025-12-31", "readOn": "2026-10-14", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/1549346/000154934626000008/sstk-20251231.htm", "sourceLabel": "Shutterstock, Form 10-K for the year ended December 31, 2025", "quote": "Our Data, Distribution, and Services revenues increased by 16%, to $203.3 million in 2025, as compared to 2024.", "addedOn": "2026-10-14" }, { "id": "troveo-content-owners-7000-50m-paid-2026", "dataType": "video", "source": "Troveo (creator, studio and archive video licensed for AI training)", "fact": "Troveo states it has paid more than $50 million to creators, has about 7,000 content owners, and has licensed more than 7 million hours of video; its homepage adds that it delivered more than 1 million hours of footage to one enterprise client, and its resource pages state 95 percent of licensors signed exclusively.", "figure": "$50M+ paid; 7K content owners; 7M+ hours licensed", "channel": "Licensing marketplace (video to AI labs)", "date": "2026-10-14", "readOn": "2026-10-14", "basis": "stated", "url": "https://www.troveo.ai/content-owners", "sourceLabel": "Troveo, For Content Owners page", "quote": "$50M+ ... paid To creators globally ... of content owners Earning passive income ... hours Video content licensed", "addedOn": "2026-10-14" }, { "id": "wirestock-40m-run-rate-15m-paid-700k-creators-2026", "dataType": "images", "source": "Wirestock (photos, video, design and 3D assets made to order for AI labs)", "fact": "Wirestock raised a $23 million Series A in May 2026 after pivoting from stock distribution to AI training data; it sources images, videos, design assets and 3D content from more than 700,000 artists and designers, supplies six of the largest foundation model developers, runs at $40 million in annual revenue and has paid $15 million to contributors since 2023.", "figure": "$40 million annual run-rate revenue; $15 million paid to contributors since 2023; 700,000+ artists", "channel": "Creator marketplace (made-to-order visual data for AI labs)", "date": "2026-05-19", "readOn": "2026-10-14", "basis": "reported", "url": "https://theaiinsider.tech/2026/05/19/ai-training-data-marketplace-wirestock-closes-23m-funding-round-after-pivoting-from-stock-photography-to-multimodal-data-supply/", "sourceLabel": "The AI Insider, Wirestock closes $23M funding round after pivoting to multimodal data supply", "quote": "Wirestock currently generates $40 million in annual run-rate revenue and has paid out $15 million to contributors since the pivot began in 2023.", "addedOn": "2026-10-14" }, { "id": "kled-3-million-files-uploaded-daily-2026", "dataType": "images", "source": "Kled (consumer app paying people for their existing photos and files)", "fact": "Kled's operator, Nitrility Inc., claimed in early 2026 a $100 million valuation and 3 million files uploaded daily by consumers selling their camera rolls for AI training.", "figure": "3 million files uploaded daily (company claim)", "channel": "Consumer app (camera rolls sold for AI training)", "date": "2026-02-03", "readOn": "2026-10-14", "basis": "reported", "url": "https://flexswag.us/2026/02/03/kled-app-sell-your-pics-for-ai-training-is-it-worth-it/", "sourceLabel": "Flexswag, Kled App: sell your pics for AI training, is it worth it?", "quote": "The company (Nitrility Inc.) claims a $100M valuation, 3 million files uploaded daily", "addedOn": "2026-10-14" }, { "id": "informa-taylor-francis-ai-partnerships-2024", "dataType": "text-archives", "source": "Taylor & Francis (Informa) archive content licensed to AI companies", "fact": "Informa reported that its AI partnerships generated more than $75 million of non-recurring data access revenue for Taylor & Francis in 2024; its 2025 report says this was not expected to repeat to the same degree in 2025.", "figure": "$75m+", "channel": "Direct licences from the publisher to AI technology companies", "date": "2025-03", "readOn": "2026-10-14", "basis": "stated", "url": "https://www.informa.com/globalassets/documents/investor-relations/2025/2024-informa-annual-report.pdf", "sourceLabel": "Informa Annual Report and Accounts 2024, Strategic Report", "quote": "Our AI partnerships generated $75m+ of non-recurring data access revenue for Taylor & Francis in 2024, further diversifying our growth.", "addedOn": "2026-10-14" }, { "id": "wiley-q1-fy2027-ai-license-revenue", "dataType": "text-archives", "source": "Wiley research and learning content licensed for AI (including content Wiley licenses from other publishers)", "fact": "Wiley's AI license revenue was $13.7 million in the three months ended 31 July 2026, against $28.9 million a year earlier; AI license revenue in its Learning (books) segment was zero, against $13.1 million.", "figure": "$13.7 million (prior year $28.9 million)", "channel": "Direct licences from the publisher, plus Research Solutions licences that include other publishers' content", "date": "2026-09-04", "readOn": "2026-10-14", "basis": "stated", "url": "https://www.sec.gov/Archives/edgar/data/107140/000162828026060632/jwa-20260731.htm", "sourceLabel": "John Wiley & Sons, Form 10-Q for the quarter ended 31 July 2026", "quote": "AI license revenue was $13.7 million for the three months ended July 31, 2026 as compared with $28.9 million in the prior year.", "addedOn": "2026-10-14" }, { "id": "spirit-archive-microsoft-365-volume", "dataType": "company-records", "source": "Spirit Airlines' bankruptcy estate: the Microsoft 365 records in the de-identified data sale", "fact": "The asset schedule of the Spirit Airlines de-identified data sale marks as included about 100 million emails across 80,000 email accounts, 17,082,644 OneDrive items, 20,577,677 SharePoint items and 500,000,000 Teams items, as the Association of Flight Attendants states in its objection (Dkt. 1489).", "figure": "about 100 million emails; 500,000,000 Teams items", "channel": "Court-supervised sale out of a Chapter 11 estate", "date": "2026-08-18", "readOn": "2026-10-15", "basis": "docket", "url": "https://chapter11cases.com/blogs/news/the-spirit-airlines-deidentified-data-sale-auction-results-objections-and-the-september-30-hearing", "sourceLabel": "Chapter11Cases (Stretto), Spirit Airlines de-identified data sale, summary of Dkt. 1489", "quote": "The Association of Flight Attendants states the schedule marks as included approximately 100 million emails across 80,000 email accounts, 17,082,644 OneDrive items, 20,577,677 SharePoint items and 500,000,000 Teams items", "addedOn": "2026-10-15" }, { "id": "depue-labs-data-spend-forecast-2030", "dataType": "company-records", "source": "Leading AI labs' spending on training data (a forecast)", "fact": "Former OpenAI researcher Will DePue predicts that by the end of the decade leading AI labs could be spending more than $100 billion a year acquiring and creating the data needed to train future models. It is one researcher's forecast, not a budget.", "figure": "more than $100 billion a year (forecast for the end of the decade)", "channel": "All routes: purchases and licences of existing data, and data made to order", "date": "2026-07-07", "readOn": "2026-10-15", "basis": "reported", "url": "https://groundlevel-ai.com/p/ai-needs-a-stargate-for-data-a-new", "sourceLabel": "Ground Level AI (Sharon Goldman), on Will DePue's 'Stargate for Data' essay", "quote": "He predicts that by the end of the decade, leading AI labs could be spending more than $100 billion a year acquiring and creating the data needed to train future models.", "addedOn": "2026-10-15" }, { "id": "chinese-labs-spend-with-us-data-vendors", "dataType": "expert-work", "source": "The six largest Chinese AI labs, buying from American data-labelling companies", "fact": "The six largest Chinese AI labs spend about $500 million a year with American data-labelling companies, according to two founders in the sector who received market-size estimates from Tencent and ByteDance executives (Forbes).", "figure": "about $500 million a year", "channel": "US data vendors selling to Chinese labs", "date": "2026-08-05", "readOn": "2026-10-15", "basis": "reported", "url": "https://forbes.co/ia/otro-problema-de-silicon-valley-en-china-esta-entrenando-su-ia", "sourceLabel": "Forbes Colombia, Spanish edition of the Forbes article by Anna Tong (5 August 2026)", "quote": "En total, los seis principales laboratorios de IA de China gastan cerca de US$500 millones al año con compañías estadounidenses de etiquetado de datos", "addedOn": "2026-10-15" }, { "id": "common-paper-csa-ai-training-prohibitions-2026", "dataType": "company-records", "source": "Software customers' contracts with their vendors (Common Paper standard cloud service agreements)", "fact": "Across 16,140 signed agreements on Common Paper's standard forms, language prohibiting AI training on customer data appeared in under 1% of cloud service agreements in 2024, 11% in 2025 and 14% in 2026.", "figure": "14% of cloud service agreements in 2026 (under 1% in 2024), across 16,140 agreements", "channel": "Customer contracts that govern data a company holds about its customers", "date": "2026-07-15", "readOn": "2026-10-15", "basis": "stated", "url": "https://commonpaper.com/resources/2026-saas-contract-benchmark-report/", "sourceLabel": "Common Paper, 2026 SaaS contract benchmark report", "quote": "Prohibiting AI Training language went from showing up in under 1% of all CSAs in 2024 to 11% in 2025 and 14% in 2026 ... This report is based on 16,140 signed agreements", "addedOn": "2026-10-15" }, { "id": "common-paper-csa-ai-mentions-2026", "dataType": "company-records", "source": "Software customers' contracts with their vendors (Common Paper standard cloud service agreements)", "fact": "Language mentioning AI appeared in 53% of the cloud service agreements signed on Common Paper's standard form in the year ending June 2026, up from 47% in 2025 and 38% in 2024.", "figure": "53% (year ending June 2026)", "channel": "Customer contracts that govern data a company holds about its customers", "date": "2026-07-15", "readOn": "2026-10-15", "basis": "stated", "url": "https://commonpaper.com/resources/2026-saas-contract-benchmark-report/", "sourceLabel": "Common Paper, 2026 SaaS contract benchmark report", "quote": "Language mentioning AI appeared in 53% of CSAs signed in the year ending June 2026, up from 47% in 2025 and 38% in 2024.", "addedOn": "2026-10-15" }, { "id": "simpleclosure-h1-2026-shutdown-median-cash", "dataType": "company-records", "source": "Startups that shut down through SimpleClosure, January to June 2026", "fact": "Non-AI companies that shut down through SimpleClosure in the first half of 2026 had a median of $13,000 in the bank when they stopped; AI companies had a median of $30,000.", "figure": "$13,000 median (non-AI); $30,000 median (AI)", "channel": "Wind-down services, whose closing customers list their archives for sale", "date": "2026-08-20", "readOn": "2026-10-15", "basis": "stated", "url": "https://simpleclosure.com/blog/insights/state-of-shutdowns-h1-2026", "sourceLabel": "SimpleClosure, The H1 2026 Shutdown Report", "quote": "AI companies stopped with a median of $30,000 in the bank, more than double the $13,000 median for the non-AI cohort.", "addedOn": "2026-10-15" }, { "id": "rl-environment-six-month-half-life", "dataType": "datasets-environments", "source": "RL environments sold to AI labs", "fact": "An RL environment is often a one-time delivery with a half-life of about six months, and the same customer rarely buys the same environment twice, according to an investor's analysis of the AI data category.", "figure": "about six months (half-life)", "channel": "Environment vendors selling to frontier labs", "date": "2026-08-20", "readOn": "2026-10-15", "basis": "reported", "url": "https://nextsignalprediction.substack.com/p/data-is-a-great-place-to-start-an", "sourceLabel": "Next Signal, Data is a great place to start an AI company", "quote": "An environment is often a one-time delivery with a ~six-month half-life; the same customer rarely buys the same environment twice.", "addedOn": "2026-10-15" }, { "id": "character-ai-estate-user-chats-no-data-buyer", "dataType": "text-archives", "source": "Character Technologies (Character.AI) estate, Chapter 11, D. Del. Case 26-11603", "fact": "Character.AI filed for Chapter 11 on 12 October 2026 to sell all or substantially all assets; Disney's $20M stalking-horse bid is a non-exclusive licence to certain IP and claims and, per the reporting, excludes the app, brand, user data and conversation logs, which stay in the estate. A Kentucky complaint attached to a 40-state AG filing (Dkt. 33) counts 21.6 million US accounts by March 2025, 2.1 million declaring an age under 18. Bid deadline 20 November, auction 23 November, sale hearing 24 November 2026. The AGs ask that no buyer may train AI on chat content beyond users' consent, so the logs are unlikely to be saleable as training data.", "figure": "21.6 million US user accounts", "channel": "Chapter 11 section 363 sale (claims agent Stretto); consumer data under multistate AG objection", "date": "2026-10-15", "readOn": "2026-10-15", "basis": "reported", "url": "https://www.techtimes.com/articles/329052/20261015/characterai-bankruptcy-puts-intimate-chats-21m-users-auction-block-40-ags-warn.htm", "sourceLabel": "Tech Times, Character.AI bankruptcy puts intimate chats of 21M users on auction block, 40 AGs warn", "quote": "more than 21.6 million US user accounts had been created by March 2025, including more than 2.1 million accounts with a declared age under 18.", "addedOn": "2026-10-16" }, { "id": "halluminate-four-of-top-five-us-labs-pay", "dataType": "environment", "source": "Halluminate (RL environments and data for knowledge work, nine people, San Francisco)", "fact": "Halluminate, which builds finance and knowledge-work RL environments, says four of the top five closed-source US AI labs are paying customers, at a mid-eight-figure annualised revenue run rate, after a $30M Series A led by Oak HC/FT.", "figure": "four of the top five closed-source U.S. AI labs", "channel": "Environments built to order and sold directly to frontier labs", "date": "2026-10-01", "readOn": "2026-10-15", "basis": "reported", "url": "https://fortune.com/2026/10/01/halluminate-raises-30-million-series-a-oakhc-ft/", "sourceLabel": "Fortune, Exclusive: Nine-person Halluminate raises $30 million, counts four top U.S. AI labs as customers (Wen Shao)", "quote": "four of the top five closed-source U.S. AI labs are paying customers", "addedOn": "2026-10-16" }, { "id": "anthropic-most-aggressive-buyer-of-coding-rl-data", "dataType": "environment", "source": "Frontier labs buying RL coding tasks and environments", "fact": "SemiAnalysis reports Anthropic as the most aggressive lab in buying coding data from RL environment startups, while its own engineers have also made coding tasks for over a year; labs pay $5,000+ for one decent coding task.", "figure": "most aggressive lab by far", "channel": "RL environment startups selling coding tasks to labs", "date": "2026-07-09", "readOn": "2026-10-15", "basis": "reported", "url": "https://newsletter.semianalysis.com/p/the-future-of-meta-superintelligence", "sourceLabel": "SemiAnalysis, The Future of Meta Superintelligence: A 1 Year Progress Update", "quote": "Anthropic has been the most aggressive lab by far when it comes to buying coding data from RL environment startups", "addedOn": "2026-10-16" }, { "id": "haitian-ruisheng-owned-speech-dataset-licences-2020", "dataType": "audio-speech", "source": "Off-the-shelf speech datasets licensed by their producer (Haitian Ruisheng, China)", "fact": "In 2020 Haitian Ruisheng licensed 199,980 hours of its own off-the-shelf speech datasets (198,951 hours of speech recognition data, 1,029 hours of speech synthesis data), up from 136,685 hours in 2019 and 115,732 in 2018; its catalogue held 762 datasets at the end of 2020.", "figure": "199,980 hours licensed in 2020", "channel": "Direct non-exclusive licensing by a data vendor to tech firms in China and abroad", "date": "2021-08-09", "readOn": "2026-10-16", "basis": "stated", "url": "https://static.cninfo.com.cn/finalpage/2021-08-09/1210701086.PDF", "sourceLabel": "Haitian Ruisheng IPO prospectus (招股说明书), 9 Aug 2021", "quote": "合计 33,445 199,980 482 27,303 136,685 598 19,917 115,732 520", "addedOn": "2026-10-16" }, { "id": "bloomsbury-14087-titles-in-bartz-works-list", "dataType": "text-archives", "source": "Bloomsbury Publishing plc", "fact": "Bloomsbury says the court listed 14,087 Bloomsbury titles in the Bartz v. Anthropic settlement, at about $3,000 a work before fees, divided equally between author and publisher, paid over several instalments from the second half of its financial year.", "figure": "14,087 titles", "channel": "Litigation settlement", "date": "2026-07-22", "readOn": "2026-10-15", "basis": "stated", "url": "https://www.investegate.co.uk/announcement/rns/bloomsbury-publishing--bmy/-1-5bn-anthropic-settlement-approved/9680717", "sourceLabel": "Bloomsbury Publishing RNS, $1.5bn Anthropic Settlement Approved (via Investegate)", "quote": "The Court has listed 14,087 Bloomsbury titles as within the settlement", "addedOn": "2026-10-16" }, { "id": "cambridge-university-press-45000-books", "dataType": "text-archives", "source": "Cambridge University Press & Assessment", "fact": "Cambridge University Press has about 45,000 books available in print and digitally, the pool it is asking authors to opt in to AI licensing; most agreements run a 2 to 5 year training term.", "figure": "about 45,000 books", "channel": "Direct licence from publishers and authors (opt-in)", "date": "2025-03-17", "readOn": "2026-10-15", "basis": "reported", "url": "https://authorsalliance.org/2025/03/17/ai-licensing-an-interview-with-ben-denne-of-cambridge-university-press", "sourceLabel": "Authors Alliance, AI Licensing: An Interview with Ben Denne of Cambridge University Press", "quote": "about 45,000 books available in print and digitally at the moment", "addedOn": "2026-10-16" }, { "id": "arcadia-publishing-17000-title-catalogue", "dataType": "text-archives", "source": "Arcadia Publishing (via Slate)", "fact": "Arcadia Publishing, the US local-history publisher (History Press, Belt, Pelican imprints), has a catalogue of nearly 17,000 titles and sublicensed some of them to an unnamed technology company for AI training in two rounds in 2025.", "figure": "nearly 17,000 titles", "channel": "Publisher sublicence, author opt-out", "date": "2025-11-03", "readOn": "2026-10-15", "basis": "reported", "url": "https://slate.com/technology/2025/11/arcadia-publishing-artificial-intelligence-history-books-authors.html", "sourceLabel": "Slate, Arcadia Publishing: Your town's local history books might be used to train A.I.", "quote": "The branding remains distinct across its nearly 17,000-title catalog", "addedOn": "2026-10-16" }, { "id": "gopro-ai-training-500k-hours-submitted-2025q4", "dataType": "video", "source": "GoPro subscribers (AI Training programme, US cloud subscribers who opted in)", "fact": "GoPro's AI Training programme, launched in Q3 2025, had received more than 500,000 hours of subscriber video by its Q4 2025 report; subscribers earn 50% of the licence revenue GoPro expects to generate on their behalf, revenue was to be recognised from Q1 2026 and first payouts made later in 2026. No dollar figure has been disclosed (the Q2 2026 10-Q names the programme only).", "figure": "500,000+ hours submitted", "channel": "Device maker's cloud library licensed to AI developers with a 50% revenue share to subscribers", "date": "2026-02", "readOn": "2026-10-15", "basis": "stated", "url": "https://s21.q4cdn.com/291350743/files/doc_financials/2025/q4/GPRO-2025-12-31-Management-Commentary.pdf", "sourceLabel": "GoPro, Q4 and full-year 2025 management commentary", "quote": "500,000 hours of diverse, high-quality video content submitted to date", "addedOn": "2026-10-16" }, { "id": "veritone-data-refinery-bookings-pipeline-40m-q3-2025", "dataType": "video", "source": "Veritone Data Refinery (rights holders' archive video and audio: sports federations, broadcasters)", "fact": "Veritone cites qualified bookings and a near-term pipeline for its Data Refinery valued at $40M at the end of Q3 2025, up 400% over Q1 2025; in one instance it processed 500,000 hours of video to deliver 26,000 hours for model training.", "figure": "$40M qualified bookings and near-term pipeline", "channel": "Broker licensing rights-cleared archives to hyperscalers and model developers on a revenue share", "date": "2026-01-29", "readOn": "2026-10-15", "basis": "reported", "url": "https://uk.themedialeader.com/tv-rights-owners-promised-new-revenue-stream-if-they-license-content-for-ai-training/", "sourceLabel": "The Media Leader, TV rights owners promised new revenue stream by licensing content for AI training (29 Jan 2026)", "quote": "citing qualified bookings and a near-term pipeline valued at $40M as of the end of Q3/25", "addedOn": "2026-10-16" }, { "id": "human-motion-data-recorder-38000-hours-3900-contributors", "dataType": "video", "source": "Human Motion Data Recorder app contributors (household-chore egocentric video, mostly the Philippines)", "fact": "Launched in April 2026 to fill one robotics company's need for 50,000 hours of egocentric video, the app had collected 38,000 hours from 3,900 contributors by mid-summer 2026, paying US$3 per approved hour.", "figure": "38,000 hours from 3,900 contributors", "channel": "Phone app paying gig workers per approved hour", "date": "2026-10-09", "readOn": "2026-10-16", "basis": "reported", "url": "https://www.theglobeandmail.com/business/article-startups-are-hiring-low-paid-gig-workers-to-train-robots-on-manual/", "sourceLabel": "The Globe and Mail, Startups are hiring low-paid gig workers to train robots on manual labour (9 Oct 2026)", "quote": "by mid-summer had collected 38,000 hours of household chores from 3,900 contributors", "addedOn": "2026-10-16" } ] }