{
  "dataset": "Buyers of AI training data",
  "description": "Organisations with a public way to sell them data: what each buys, its published prices, eligibility, payment terms and introduction fees.",
  "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": "micro1",
      "name": "micro1",
      "domain": "micro1.ai",
      "kind": "principal",
      "what": "Licenses a company's operating records and turns them into training environments it sells to AI labs.",
      "wants": "SOPs, knowledge bases, internal documentation, CRM data, project histories, QA processes, decision records.",
      "pays": "$100k+ for a qualified partnership, $500k+ for large or multi-team datasets, $1M+ for highly unique data.",
      "low": 100000,
      "high": 1000000,
      "paidWhen": "Not published.",
      "scrub": "micro1 removes personal information \"when relevant\" and may rewrite some datasets synthetically. Companies keep ownership.",
      "criteria": {
        "statuses": [
          "operating"
        ],
        "minEmployees": 30,
        "regions": [
          "us",
          "uk-ca",
          "eu",
          "other-english"
        ],
        "english": true
      },
      "eligibility": "Operationally mature companies with 30+ employees that work and document in English. US first, then the UK, Canada and other Western markets. A separate track (its posting of 8 Oct 2026) takes companies that have shut down, with 50+ employees at their peak and substantial Teams or Slack history.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "Up to $50,000 per referred company (up to $25,000 at the programme's launch in June 2026), 7.5% of the licence price where no amount is posted, and a referral field of $75,000 on its posting for shut-down companies; a partner who recruits another partner earns 20% of that partner's payout on top.",
      "introducerTerms": {
        "rate": 0.075,
        "cap": 50000
      },
      "record": {
        "tier": "confirmed",
        "says": "Series B priced at a $4 billion valuation, led by Dream Ventures with a roughly $25 million employee tender, a run rate above $1 billion with about 180 people (micro1, 12 Oct 2026), and a $1 billion, 12-month commitment to buy company data financed with capital from Citi and Hercules Capital (micro1, 9 Oct 2026; size and form not stated). Hercules Capital's own quarterly filing (10-Q, 30 Jun 2026) lists a $10.875 million senior secured loan to micro1, due June 2029, made before that announcement. Its own filing in the Spirit Airlines bankruptcy states more than 50 data transactions in 45 days (Fortune, 14 Sep 2026).",
        "source": "https://www.micro1.ai/blog/q3",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.micro1.ai/data-partnerships",
      "sourceLabel": "micro1, data partnerships page",
      "readOn": "2026-10-11"
    },
    {
      "id": "handshake",
      "name": "Handshake AI",
      "domain": "joinhandshake.com",
      "kind": "principal",
      "what": "A supplier of expert work to AI labs that also buys operational data from businesses in every major industry and delivers it to labs.",
      "wants": "Operational data from the tools a business runs on (Gmail, Slack, Drive, NetSuite, QuickBooks, Figma and more), transferred as it exists today.",
      "pays": "\"Payouts range from $100K to $4M.\" Its estimator shows $383,083 to $896,969 for 100 employees and 10 years.",
      "low": 100000,
      "high": 4000000,
      "paidWhen": "\"Payment terms are negotiated as part of your agreement and tied to defined delivery milestones.\" It says an initial review and pricing come within 12 to 24 hours of the details it needs, and the whole process typically takes 7 to 14 business days (FAQ on its data-partnerships page, 17 Oct 2026).",
      "scrub": "Handshake removes direct and indirect business and personal identifiers before it uses the data, and says no de-identification removes the risk of re-identification entirely. Its page no longer says who carries the legal risk of the transfer (17 Oct 2026), so the contract has to.",
      "criteria": {
        "statuses": [
          "operating"
        ],
        "minEmployees": 20,
        "minYears": 3
      },
      "eligibility": "\"Companies with 20+ full time employees & 3+ years in operation\", in any industry.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "\"You can earn $10K for each successful referral that becomes a Handshake AI data partner and completes its data transfer\" (FAQ on its data-partnerships page, read 17 Oct 2026).",
      "introducerTerms": {
        "flat": 10000
      },
      "record": {
        "tier": "confirmed",
        "says": "Its AI unit passed $1 billion of gross annualised revenue in 16 months (TIME, 27 Aug 2026), and it bought Cleanlab in January 2026 (TechCrunch, 28 Jan 2026). Its own page says \"Payments to date have ranged from $100K to $4M, depending on the scope of the partnership\" (17 Oct 2026), but no seller has confirmed a sale of company records to it.",
        "source": "https://time.com/collection/time100-ai/2026/garrett-lord/",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://joinhandshake.com/ai/data-partnerships/license",
      "sourceLabel": "Handshake AI, data licensing page",
      "readOn": "2026-10-15"
    },
    {
      "id": "scale",
      "name": "Scale AI",
      "domain": "scale.com",
      "kind": "principal",
      "what": "A long-standing data vendor to AI labs that licenses structured, de-identified records from companies' everyday systems.",
      "wants": "Structured, de-identified records from data warehouses, chat, email, documents, work tools and business platforms.",
      "pays": "\"$10K–$1M+ illustrative value per data partnership\", scaling with how often data is delivered.",
      "low": 10000,
      "high": 1000000,
      "paidWhen": "Not published. A scoping call first; a partnership can begin with a very limited scope.",
      "scrub": "Scale anonymises the data inside the company's own environment before extraction, and each package needs the company's approval.",
      "criteria": {
        "statuses": [
          "operating"
        ],
        "minEmployees": 40
      },
      "eligibility": "\"40+ employees is our preferred starting point\", assessed case by case. Established systems and real operational workflows.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "Up to $50,000, paid only if Scale signs a partnership agreement.",
      "introducerTerms": {
        "flat": 50000,
        "upTo": true
      },
      "record": {
        "tier": "confirmed",
        "says": "Meta paid about $14.3 billion for 49 percent of it (TechCrunch, 13 Jun 2025). Its programme for companies' records is one page, and no seller has confirmed a deal through it.",
        "source": "https://techcrunch.com/2025/06/13/scale-ai-confirms-significant-investment-from-meta-says-ceo-alexandr-wang-is-leaving/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://scale.com/data-partnership",
      "sourceLabel": "Scale AI, data partnership page",
      "readOn": "2026-10-15"
    },
    {
      "id": "mercor",
      "name": "Mercor",
      "domain": "mercor.com",
      "kind": "principal",
      "what": "Connects to a company's tools, extracts and anonymises the records with its own automated pipeline, and licenses the dataset to AI labs.",
      "wants": "Workflow data from Slack, Teams, Gmail, Outlook, Confluence, Salesforce, GitHub, Drive, Notion and 100+ other tools.",
      "pays": "No figures. \"Compensation is based on data volume, the tools you connect, and the depth of your data.\"",
      "low": null,
      "high": null,
      "paidWhen": "By wire on delivery of the anonymised dataset, typically within 2 to 4 weeks of the signed agreement.",
      "scrub": "Automated extraction and anonymisation, paid for by Mercor. It says no buyer sees raw data.",
      "criteria": {
        "statuses": [
          "operating"
        ],
        "industries": [
          "Legal",
          "Healthcare",
          "Finance and insurance",
          "Professional services",
          "Manufacturing",
          "Retail and e-commerce",
          "Real estate",
          "Logistics and supply chain"
        ]
      },
      "eligibility": "No size, region or language rule published. It names eight industries it is looking for.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": "5% of the deal, up to $100,000, paid after the company's first delivery (shown in its referral app; the public page names no amount).",
      "introducerTerms": {
        "rate": 0.05,
        "cap": 100000
      },
      "record": {
        "tier": "confirmed",
        "says": "$350 million Series C at a $10 billion valuation led by Felicis (TechCrunch, 27 Oct 2025). In July 2026 its CEO put the run rate above $2 billion, it was reported in talks to raise at $20 billion with a term sheet (TechCrunch citing Bloomberg, 9 Jul 2026; no closed round found since), and it announced buying Deeptune, a company that helps train AI agents (price not disclosed). It is the $7.5 million backup bidder for Spirit Airlines' data in court. No seller has yet confirmed a sale of company records to it.",
        "source": "https://techcrunch.com/2026/07/09/mercor-is-in-talks-for-a-20b-valuation/",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://mercor.com/data",
      "sourceLabel": "Mercor, data page",
      "readOn": "2026-10-11"
    },
    {
      "id": "superannotate",
      "name": "SuperAnnotate",
      "domain": "superannotate.com",
      "kind": "principal",
      "what": "A data annotation platform that prepares businesses' operational records and licenses them to AI labs and enterprise AI teams the seller approves.",
      "wants": "Communications, process documents, decision trails, domain records and screen recordings.",
      "pays": "\"Indicative $100K – $1M+\", priced on rarity, depth and exclusivity.",
      "low": 100000,
      "high": 1000000,
      "paidWhen": "\"Payment on delivery, and on each relicense.\"",
      "scrub": "SuperAnnotate removes identifying detail to the standard the seller's team sets, and deletes on request.",
      "criteria": {},
      "eligibility": "Any size, anywhere, in any industry, with records built up over time.",
      "takes": [
        "company-records",
        "expert-work"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "$36 million Series B led by Socium Ventures with NVIDIA (InTech, Nov 2024). Its programme for companies' records is one page with no deal anyone confirms.",
        "source": "https://intech.am/superannotate-secures-36m-in-series-b-funding-round/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.superannotate.com/data-partnerships",
      "sourceLabel": "SuperAnnotate, data partnerships page",
      "readOn": "2026-10-15"
    },
    {
      "id": "turing",
      "name": "Turing (Lazarus)",
      "domain": "turing.com",
      "kind": "principal",
      "what": "Buys defined rights to a company's working record and its codebases, for one payment, to train frontier models.",
      "wants": "Team communication, source code with reviews and tickets, wikis, runbooks, schemas and the reasoning behind decisions.",
      "pays": "\"A single payout of up to $1M\" for an operating record. For code repositories its process page (lazarus.turing.com/process, read 16 Oct 2026) says \"Repos typically pay $10k–$100k+ depending on quality and complexity\".",
      "low": 10000,
      "high": 1000000,
      "paidWhen": "One payment after a sample review and a final offer. Intake call within one business day under NDA.",
      "scrub": "Turing removes personal data and secrets in an air-gapped environment, on its side.",
      "criteria": {
        "statuses": [
          "operating",
          "closing",
          "closed"
        ],
        "code": {
          "minLinesPerRepo": 50000,
          "humanWritten": true,
          "withTests": true
        }
      },
      "eligibility": "Operating, winding-down and closed companies, strongest demand in the US, UK, Canada and EU. Codebases: 50,000+ lines, human-written, with tests.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": "Up to $10,000 per referral.",
      "introducerTerms": {
        "flat": 10000,
        "upTo": true
      },
      "record": {
        "tier": "confirmed",
        "says": "$111 million Series E led by Khazanah Nasional (TechCrunch, 6 Mar 2025). No seller has confirmed a Project Lazarus deal.",
        "source": "https://techcrunch.com/2025/03/06/turing-a-key-coding-provider-for-openai-and-other-llm-producers-raises-111m-at-a-2-2b-valuation",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://lazarus.turing.com/",
      "sourceLabel": "Turing, Lazarus page",
      "readOn": "2026-10-11"
    },
    {
      "id": "nyne",
      "name": "Nyne",
      "domain": "nyne.ai",
      "kind": "principal",
      "what": "Licenses the accounting, customer and scheduling records of established US businesses.",
      "wants": "QuickBooks, Xero and NetSuite records, Salesforce or HubSpot CRM, support tickets, jobs and dispatch data.",
      "pays": "\"Partnerships are commonly worth $100K or more.\" It says it will beat a qualifying written offer for the same data.",
      "low": 100000,
      "high": null,
      "paidWhen": "Not published. A 15-minute scoping call comes first.",
      "scrub": "Nyne de-identifies before release and lets the company exclude whole categories.",
      "criteria": {
        "statuses": [
          "operating"
        ],
        "minEmployees": 20,
        "minYears": 3,
        "regions": [
          "us"
        ]
      },
      "eligibility": "At least 20 full-time US W-2 employees and three years in business.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "Not stated on its sell-company-data page (read 16 Oct 2026); up to $20,000 per company introduced, by email to its CEO, as reported earlier, is unconfirmed.",
      "record": {
        "tier": "confirmed",
        "says": "$5.3 million seed led by Wischoff Ventures and South Park Commons, founders named (TechCrunch, 13 Mar 2026). Buying company records is a new line, and no seller has confirmed a deal.",
        "source": "https://techcrunch.com/2026/03/13/nyne-founded-by-a-father-son-duo-gives-ai-agents-the-human-context-theyre-missing/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://nyne.ai/sell-company-data/",
      "sourceLabel": "Nyne, sell company data page",
      "readOn": "2026-10-11"
    },
    {
      "id": "ooak",
      "name": "Ooak",
      "domain": "ooakdata.com",
      "kind": "principal",
      "what": "Anonymises a company's records into a code-named digital twin and turns it into training environments for AI labs.",
      "wants": "Documents, communications and tool data from Gmail, Slack, Notion, Jira, Drive, SharePoint and Teams.",
      "pays": "\"Earn up to $1M\" (raised from $300K by 15 Oct 2026). Deals to date run from $20K to several hundred thousand dollars.",
      "low": 20000,
      "high": 1000000,
      "paidWhen": "2 to 4 weeks, paid to the company, not to individuals.",
      "scrub": "Ooak strips personal data with its own engine and pays for the work; the raw archive is deleted within 30 days with written confirmation.",
      "criteria": {
        "statuses": [
          "operating",
          "closing",
          "closed"
        ],
        "minEmployees": 20
      },
      "eligibility": "Companies of 20+ employees, operating, winding down or closed. A liquidator or administrator can sign.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "partial",
        "says": "A Y Combinator company (Summer 2026) with its founders named in YC's directory. Its revenue and its 20 companies are its own statements; no seller confirms a deal.",
        "source": "https://www.ycombinator.com/companies/ooak-data",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.ooakdata.com/partner",
      "sourceLabel": "Ooak, partner page",
      "readOn": "2026-10-15"
    },
    {
      "id": "troveo",
      "name": "Troveo",
      "domain": "troveo.ai",
      "kind": "marketplace",
      "what": "Licenses non-public data on its owners' behalf (creators' video, audio and game libraries; companies' records) and pays a share of every sale.",
      "wants": "Company exports (email, chat, CRM, tickets, files); 1080p or 4K footage; audio and gameplay libraries.",
      "pays": "\"AI labs and startups are actively paying six figures for company data exports\", and Troveo pays the holder \"on every sale, not just the first\".",
      "low": 100000,
      "high": null,
      "paidWhen": "For in-demand datasets, signed deal to first payout in under a month.",
      "scrub": "Troveo scrubs names, contact details, credentials and client identities to a documented standard, at no charge.",
      "criteria": {},
      "eligibility": "No minimum stated; value scales with team size. Content must be owned or rights-held.",
      "takes": [
        "company-records",
        "video",
        "audio-speech",
        "images",
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "Named creators and a talent manager confirm its payouts for video libraries (Semafor, 30 Oct 2025); it states \"$20M+ Paid to Licensors\" and \"40+ Active Buyers\" (read 16 Oct 2026) and 7,000+ data providers in its network. Its only disclosed round is a $4.5 million seed led by Seven Seven Six (Nov 2024). Its line for companies' records is newer and unconfirmed.",
        "source": "https://www.troveo.ai/business-data",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.troveo.ai/business-data",
      "sourceLabel": "Troveo, business data page",
      "readOn": "2026-10-14"
    },
    {
      "id": "protege",
      "name": "Protege",
      "domain": "withprotege.ai",
      "kind": "marketplace",
      "what": "A licensing platform that links data from many providers into datasets for AI model builders and pays each provider a revenue share.",
      "wants": "Healthcare records and imaging, video, audio and speech, spatial and physical data, and other domains.",
      "pays": "A revenue share per deal, by the agreed partnership terms.",
      "low": null,
      "high": null,
      "paidWhen": "Revenue share payments when the data is included in a qualified deal.",
      "scrub": "Protege de-identifies, certifies and tokenises health data with the provider; automated identifier removal elsewhere.",
      "criteria": {},
      "eligibility": "Providers who own their data outright or control it under broad rights.",
      "takes": [
        "text-archives",
        "video",
        "audio-speech",
        "images",
        "physical-world"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "A $30 million Series A extension led by a16z (8 Jan 2026; $65 million in all) on a $25 million Series A led by Footwork (13 Aug 2025), by its own releases; it bought Calliope Networks in December 2024. Partners OneMedNet and VARSAV disclosed their agreements with it.",
        "source": "https://withprotege.ai/articles/news/protege-a16z-30million-fundraise/",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.withprotege.ai/data-providers",
      "sourceLabel": "Protege, data providers page",
      "readOn": "2026-10-11"
    },
    {
      "id": "defined-ai",
      "name": "Defined.ai",
      "domain": "defined.ai",
      "kind": "marketplace",
      "what": "An AI data marketplace whose partner programme lists businesses' datasets and pays licensing fees from each sale.",
      "wants": "Audio, video, image, text and multimodal datasets; in high demand: cinematic video, anime, coding, STEM.",
      "pays": "No rate card. It states \"$600K+ average partner revenue\" and \"$1M+\" for five or more partners.",
      "low": null,
      "high": null,
      "paidWhen": null,
      "scrub": "The seller must have consent and the right to sell; Defined.ai reviews rights and compliance.",
      "criteria": {},
      "eligibility": "Any business that makes or aggregates data, with no fixed minimum volume.",
      "takes": [
        "audio-speech",
        "video",
        "images",
        "text-archives",
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "Reported 65% revenue growth for 2025 and 143% net revenue retention (27 Jan 2026); it says it has raised over $85 million, its last disclosed round a $50.5 million Series B in 2021 as DefinedCrowd.",
        "source": "https://defined.ai/press-room/defined-ai-reports-sixty-five-percent-revenue-growth",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.defined.ai/partnerships",
      "sourceLabel": "Defined.ai, partnerships page",
      "readOn": "2026-10-11"
    },
    {
      "id": "sieve",
      "name": "Sieve",
      "domain": "sievedata.com",
      "kind": "principal",
      "what": "A multimodal data lab that licenses existing video, audio and image libraries and pays businesses to record their teams' physical work.",
      "wants": "Existing video, audio, image and interactive libraries; recordings of physical work in kitchens, factories, studios and labs.",
      "pays": "Pays after delivery is checked (\"Sieve Checks them, then pays\"); recording is paid per hour and no rate is published.",
      "low": null,
      "high": null,
      "paidWhen": "After the files are delivered and validated.",
      "scrub": "Sieve excludes personal material, and everyone recorded gives informed consent.",
      "criteria": {},
      "eligibility": "Libraries from studios, broadcasters, publishers, universities and game companies. Its recording programme starts at 50 staff and 500 hours a week.",
      "takes": [
        "video",
        "audio-speech",
        "images",
        "physical-world",
        "datasets-environments"
      ],
      "introducerFee": "\"Varies by partnership\".",
      "record": {
        "tier": "partial",
        "says": "Its job postings name Matrix Partners, Swift Ventures, Y Combinator and AI Grant as investors in a Series A of undisclosed size; the $15.5 million figure seen elsewhere has no primary source. It buys media and recorded work, not company records.",
        "source": "https://sieve-data.notion.site/Roles-Sieve-ad53969df1d846b4b890e2c12dd5ef91",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.sievedata.com/partners",
      "sourceLabel": "Sieve, partners page",
      "readOn": "2026-10-14"
    },
    {
      "id": "wirestock",
      "name": "Wirestock",
      "domain": "wirestock.io",
      "kind": "marketplace",
      "what": "A creator platform that licenses photographers', videographers' and illustrators' work to AI labs through paid projects and dataset deals.",
      "wants": "Photos, video and illustration the creator holds the rights to, with releases for people and property.",
      "pays": "Paid per approved asset on projects; in dataset deals, by each creator's share of the pool.",
      "low": null,
      "high": null,
      "paidWhen": "By PayPal or Payoneer once the balance reaches $30.",
      "scrub": "Creators supply model and property releases.",
      "criteria": {},
      "eligibility": "Any creator with the rights to license their uploads. Non-exclusive; creators keep copyright.",
      "takes": [
        "images",
        "video"
      ],
      "introducerFee": "10% of a referred creator's earnings, up to $25.",
      "record": {
        "tier": "confirmed",
        "says": "A $23 million Series A led by Nava Ventures, a $40 million annual run rate and $15 million paid to contributors (TechCrunch, 14 May 2026), as the company stated them.",
        "source": "https://techcrunch.com/2026/05/14/wirestock-raises-23m-to-supply-multi-modal-data-to-ai-labs/",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://wirestock.io/docs/faq",
      "sourceLabel": "Wirestock, FAQ",
      "readOn": "2026-10-11"
    },
    {
      "id": "kled",
      "name": "Kled",
      "domain": "kled.ai",
      "kind": "marketplace",
      "what": "A consumer app where individuals upload photos, video and files, and earn when the data is licensed to AI companies.",
      "wants": "Photos, video, audio and files individuals own, plus labelling tasks.",
      "pays": null,
      "low": null,
      "high": null,
      "paidWhen": "Earnings collect in a wallet that can be cashed out at any time.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Individuals, through its iOS app.",
      "takes": [
        "images",
        "video",
        "audio-speech"
      ],
      "introducerFee": null,
      "record": {
        "tier": "partial",
        "says": "A $5.5 million seed (10 Mar 2026) and $3 million from The DATA Foundation (25 Jun 2026), $14 million in all by its own count; the $10 million on its site is the total after the seed, not a round. A $12 million non-exclusive data deal and an $8 million request for financial documents are its own claims, naming no buyer.",
        "source": "https://www.kled.ai/blog/the-data-foundation-invests-3m-in-kled",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://www.kled.ai/",
      "sourceLabel": "Kled, site",
      "readOn": "2026-10-11"
    },
    {
      "id": "datarade",
      "name": "Datarade",
      "domain": "datarade.ai",
      "kind": "marketplace",
      "what": "A data marketplace where registered businesses list data products, receive buyer leads and set their own licence terms.",
      "wants": "Data products of any kind, including AI training datasets.",
      "pays": "The seller sets the price. Datarade takes 30% commission, or 20% to 15% on a paid plan of $6,000 to $12,000 a year.",
      "low": null,
      "high": null,
      "paidWhen": null,
      "scrub": "The seller removes personal data; no unanonymised personal data is permitted.",
      "criteria": {},
      "eligibility": "Registered businesses of any size.",
      "takes": [
        "datasets-environments",
        "text-archives",
        "images",
        "audio-speech",
        "video"
      ],
      "introducerFee": null,
      "basis": "stated",
      "source": "https://providers.datarade.ai/apply",
      "sourceLabel": "Datarade, provider page",
      "readOn": "2026-10-11"
    },
    {
      "id": "created-by-humans",
      "name": "Created by Humans",
      "domain": "createdbyhumans.ai",
      "kind": "marketplace",
      "what": "A rights platform for books: authors, agents and publishers opt works in, and it negotiates training and reference licences with AI developers.",
      "wants": "Books, licensed for training, reference or transformative use, work by work.",
      "pays": "Individual deals expected to run from hundreds to thousands of dollars, in proportion to a book's sales (its chief executive, reported).",
      "low": null,
      "high": null,
      "paidWhen": "Agreements are negotiated for the author's approval; payments are tracked in a dashboard.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Authors, literary agents and publishers.",
      "takes": [
        "text-archives"
      ],
      "introducerFee": null,
      "basis": "reported",
      "source": "https://www.pw.org/content/an_ai_licensing_platform_takes_off",
      "sourceLabel": "Poets & Writers",
      "readOn": "2026-10-16"
    },
    {
      "id": "halluminate",
      "name": "Halluminate",
      "domain": "halluminate.ai",
      "kind": "principal",
      "what": "Builds training environments and benchmarks for professional work, finance first, and takes interest from companies with data to license.",
      "wants": "Records of professional work, finance and professional services first, through a data-partnership interest form.",
      "pays": null,
      "low": null,
      "high": null,
      "paidWhen": null,
      "scrub": null,
      "criteria": {},
      "eligibility": "Not published.",
      "takes": [
        "company-records",
        "expert-work"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "$30 million Series A led by Oak HC/FT, with four large US labs as customers (Fortune, 1 Oct 2026). It builds training environments; buying records is a side door.",
        "source": "https://fortune.com/2026/10/01/halluminate-raises-30-million-series-a-oakhc-ft/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.halluminate.ai/data-partnerships",
      "sourceLabel": "Halluminate, data partnerships page",
      "readOn": "2026-10-11"
    },
    {
      "id": "datavendor",
      "name": "DataVendor (HUD)",
      "domain": "datavendor.ai",
      "kind": "marketplace",
      "what": "HUD's marketplace connecting builders of training environments and task sets with lab research teams.",
      "wants": "Finished training environments and task sets the seller fully owns.",
      "pays": null,
      "low": null,
      "high": null,
      "paidWhen": null,
      "scrub": null,
      "criteria": {},
      "eligibility": "Invite only; its seller terms are not public.",
      "takes": [
        "datasets-environments"
      ],
      "introducerFee": "5% of what DataVendor receives on a referred vendor's deals within 180 days, 7.5% for supply named up front and accepted in writing, up to 10% for rare supply, out of its own share.",
      "record": {
        "tier": "partial",
        "says": "DataVendor is the marketplace of HUD, whose $16 million Series A led by Standard Capital with Y Combinator is on HUD's own blog (30 Jun 2026). The marketplace itself is in beta, and no sale through it is confirmed.",
        "source": "https://www.hud.ai/blog/announcing-16m-series-a",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://datavendor.ai/",
      "sourceLabel": "DataVendor, site",
      "readOn": "2026-10-11"
    },
    {
      "id": "prime-intellect",
      "name": "Prime Intellect",
      "domain": "primeintellect.ai",
      "kind": "collector",
      "what": "An open-source AI lab that pays bounties for training environments and evaluations built for its open Environments Hub.",
      "wants": "Training environments and evaluations, released open source.",
      "pays": "$100 to $500 for open bounties; $1,000 to $5,000+ for application-only tasks.",
      "low": 100,
      "high": 5000,
      "paidWhen": "When the task is completed, including reproduction of known scores for reimplemented benchmarks.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Anyone for open tasks; experienced builders by application for larger ones.",
      "takes": [
        "datasets-environments"
      ],
      "introducerFee": null,
      "basis": "stated",
      "source": "https://www.primeintellect.ai/blog/scaling-environments-program",
      "sourceLabel": "Prime Intellect, environments programme",
      "readOn": "2026-10-11"
    },
    {
      "id": "datacurve",
      "name": "Datacurve (Shipd)",
      "domain": "shipd.ai",
      "kind": "collector",
      "what": "A coding-data vendor whose Shipd platform pays engineers bounties for solving hard coding problems that become training data.",
      "wants": "Solved coding and STEM tasks, by engineers in 40+ countries.",
      "pays": "Example bounties in the sample panels on its homepage run from $80 to $220 (read 16 Oct 2026; 2,847 registered contributors on its live counter). The panels show sample issues and challenges, not a rate card: \"New quests launch every month across dozens of STEM categories\".",
      "low": 80,
      "high": 220,
      "paidWhen": null,
      "scrub": null,
      "criteria": {},
      "eligibility": "Engineers anywhere, after vetting.",
      "takes": [
        "source-code",
        "expert-work"
      ],
      "introducerFee": null,
      "basis": "stated",
      "source": "https://shipd.ai/",
      "sourceLabel": "Shipd, site",
      "readOn": "2026-10-16"
    },
    {
      "id": "grepped",
      "name": "Grepped",
      "domain": "grepped.ai",
      "kind": "broker",
      "what": "A broker: it signs an agreement with the data owner, packages the records and sells them to AI labs.",
      "wants": "Company internals, datasets and training environments the owner has the right to sell.",
      "pays": "Its own pages cite $20K to $5M per company; its estimator shows a sample of $29K to $41.5K.",
      "low": null,
      "high": null,
      "paidWhen": "Within 7 days of the data passing review; it says most companies take two to three weeks from form to payment.",
      "scrub": "Grepped de-identifies the records itself and says no method is perfect.",
      "criteria": {},
      "eligibility": "No size or region rule published. It prefers established businesses with years of records.",
      "takes": [
        "company-records",
        "datasets-environments"
      ],
      "introducerFee": "10% of what it pays the owner, up to $100,000, plus 2% of the fees of referrers a referrer brings in. Its terms forbid a referrer from helping the owner sell or license the data any other way, for 12 months after they end.",
      "introducerTerms": {
        "rate": 0.1,
        "cap": 100000
      },
      "record": {
        "tier": "own-claims",
        "says": "Domain registered in February 2026, and until July 2026 it hosted an AI interface-component generator; its founder wrote on 5 Oct that he had entered the space two weeks earlier. No funding or named backers found. Its $48 million annual run rate is its founder's own post, and no seller or lab confirms a deal.",
        "source": "https://www.grepped.ai/business/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.grepped.ai/business/",
      "sourceLabel": "Grepped, business page",
      "readOn": "2026-10-11"
    },
    {
      "id": "fileyield",
      "name": "FileYield",
      "domain": "fileyield.com",
      "kind": "broker",
      "what": "A private brokerage that lists or introduces companies' datasets to AI labs; it does not prepare, de-identify or hold the data.",
      "wants": "Any proprietary dataset, including small ones of 10,000 to 100,000 records.",
      "pays": "No figures. It charges sellers 8% and buyers 6% commission.",
      "low": null,
      "high": null,
      "paidWhen": "Buyer and seller agree payment terms in their own contract.",
      "scrub": "The seller is responsible; FileYield does not strip personal data.",
      "criteria": {},
      "eligibility": "No minimum size.",
      "takes": [
        "company-records",
        "audio-speech",
        "video",
        "images",
        "text-archives",
        "source-code",
        "datasets-environments"
      ],
      "introducerFee": null,
      "basis": "stated",
      "source": "https://fileyield.com/sell-data",
      "sourceLabel": "FileYield, sell data page",
      "readOn": "2026-10-11"
    },
    {
      "id": "corpus-bridge",
      "name": "Corpus Bridge",
      "domain": "corpusbridge.com",
      "kind": "broker",
      "what": "An introducer that assesses a business's records and passes them to an unnamed brokerage partner, on a success fee.",
      "wants": "Records of established businesses with years of history.",
      "pays": "It cites licences of about $10,000 to several hundred thousand dollars for ordinary companies.",
      "low": 10000,
      "high": null,
      "paidWhen": "The buyer pays on contract milestones, for example half on signature and half on delivery.",
      "scrub": "Its partner removes or replaces names, contact details and account ids.",
      "criteria": {},
      "eligibility": "Established businesses; no number given.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": null,
      "basis": "stated",
      "source": "https://corpusbridge.com/",
      "sourceLabel": "Corpus Bridge, site",
      "readOn": "2026-10-16"
    },
    {
      "id": "simpleclosure",
      "name": "SimpleClosure AssetHub",
      "domain": "simpleclosure.com",
      "kind": "wind-down",
      "what": "A company-shutdown service whose AssetHub brokers the sale or licence of code, workspace data, IP and domains, for a percentage on success.",
      "wants": "Source code, and workspace data such as email, Slack, Notion and Drive.",
      "pays": "\"Tens to hundreds of thousands of dollars for unique assets\". Early deals were reported at $10,000 to $100,000.",
      "low": 10000,
      "high": null,
      "paidWhen": "Not published.",
      "scrub": "SimpleClosure processes personal data before licensing and excludes compensation, health and other sensitive records.",
      "criteria": {
        "statuses": [
          "operating",
          "closing",
          "closed"
        ]
      },
      "eligibility": "Startups shutting down, and now companies still operating too.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": "Up to $1,000 for a referred shutdown; nothing published for data sales.",
      "basis": "stated",
      "source": "https://simpleclosure.com/blog/insights/what-your-assets-are-worth/",
      "sourceLabel": "SimpleClosure, What your assets are worth",
      "readOn": "2026-10-14"
    },
    {
      "id": "sunset",
      "name": "Sunset",
      "domain": "sunsethq.com",
      "kind": "wind-down",
      "what": "A flat-fee wind-down service for venture-backed startups that also sells a closing company's anonymised data to AI labs.",
      "wants": "Anonymised records and code of a company that is closing.",
      "pays": null,
      "low": null,
      "high": null,
      "paidWhen": null,
      "scrub": null,
      "criteria": {
        "statuses": [
          "closing",
          "closed"
        ]
      },
      "eligibility": "Startups winding down.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "partial",
        "says": "A $1.45 million seed (TechCrunch, 27 Feb 2024), and its data resale is covered by MIT Technology Review (13 Oct 2026). No seller has confirmed a data sale through it.",
        "source": "https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://sunsethq.com/product",
      "sourceLabel": "Sunset, product page",
      "readOn": "2026-10-11"
    },
    {
      "id": "humxn",
      "name": "HUMXN Research Labs",
      "domain": "humxn.io",
      "kind": "principal",
      "what": "Buys records from skilled-trades and manufacturing businesses and licenses them for AI and robotics training; also pays workers for footage of their work.",
      "wants": "Construction: bids, estimates, RFIs, change orders, site photos and daily logs. Home services: service calls, quotes, job notes, technician photos and dispatch. Manufacturing and field-inspection records.",
      "pays": "\"$250K+ Where typical data deals for a single company start, with larger archives reaching the millions.\" Workers earn $20 per hour of approved footage.",
      "low": 250000,
      "high": null,
      "paidWhen": null,
      "scrub": "HUMXN checks the business owns the data, then removes customer names, addresses and phone numbers and reviews footage and photos so no one can be identified.",
      "criteria": {},
      "eligibility": "Skilled-trades and manufacturing businesses, such as plumbing, HVAC and electrical companies. No size or region rule published.",
      "takes": [
        "company-records",
        "physical-world"
      ],
      "introducerFee": "10% of the value of the data deal, up to $100,000 per deal, for deals signed within 12 months of the introduction, paid within 30 days of closing.",
      "introducerTerms": {
        "rate": 0.1,
        "cap": 100000
      },
      "record": {
        "tier": "own-claims",
        "says": "Announced on 13 Oct 2026 a programme to acquire and license $50 million of data from skilled trades, farming and manufacturing businesses: the size of its buying target, not capital behind it. It names no backer, and no seller or lab confirms a deal.",
        "source": "https://www.newsfilecorp.com/release/318713",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://humxn.io/sell-company-data",
      "sourceLabel": "HUMXN, sell company data page",
      "readOn": "2026-10-16"
    },
    {
      "id": "polyshares",
      "name": "Polyshares",
      "domain": "polyshares.com",
      "kind": "principal",
      "what": "Buys a defined, bounded licence to a company's operating records, has them anonymised by an outside firm and supplies them to AI labs it does not name. It says it is the buyer, not an agent or a marketplace.",
      "wants": "Slack, email, tickets, CRM, files and the rest of a company's working records.",
      "pays": "Typical licences of $100K to $2M, sometimes more. Its track record page lists four anonymous licences from $110,000 to $450,000.",
      "low": 100000,
      "high": 2000000,
      "paidWhen": "An offer within one to two weeks of intake; payment net 30 to 60 days after the data is approved and anonymised.",
      "scrub": "Anonymised through a third-party firm before it reaches a lab.",
      "criteria": {},
      "eligibility": "Companies with real operating history in any sector, venture-backed, middle-market or founder-owned. Its site publishes no size or region rule.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "6% of the price it pays the company, paid on top of that price, to brokers registered in its portal, as its co-founder states publicly. A company whose name or email is already in its system cannot be submitted again, so the first record keeps the credit.",
      "introducerTerms": {
        "rate": 0.06,
        "onTop": true
      },
      "record": {
        "tier": "partial",
        "says": "Its chief executive's earlier company, Roofer.com, raised a $7.5 million seed (Dallas Innovates, 30 Jan 2024). Polyshares itself shows no funding and 2 to 10 staff, its co-founder dates its start to mid-2026, and no seller confirms any of its four listed licences.",
        "source": "https://dallasinnovates.com/north-texas-based-roofer-com-raises-7-5m-for-its-drones-ai-based-modernized-roofing/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://polyshares.com/",
      "sourceLabel": "Polyshares, site and broker portal",
      "readOn": "2026-10-15"
    },
    {
      "id": "mode-bizbuydata",
      "name": "Mode (BizBuyData)",
      "domain": "bizbuydata.com",
      "kind": "principal",
      "what": "Mode Mobile, a US company that files with the SEC, buys an agreed, filtered copy of an organisation's operating records and licenses it on for AI research and training. The programme opened on 13 October 2026.",
      "wants": "Email, documents, chat, CRM, billing, calls and transcripts, project histories.",
      "pays": "\"The operating history of companies and nonprofits could be worth $100K-$5M.\"",
      "low": 100000,
      "high": 5000000,
      "paidWhen": "A nonbinding letter of intent within 14 days; once the final offer is accepted Mode collects and de-identifies the data over 15 to 30 days, and its launch release expects payment in about 60 days.",
      "scrub": "Mode de-identifies the copy itself before it is licensed on.",
      "criteria": {},
      "eligibility": "Companies and nonprofits with several years of records they own, teams primarily in the US: 20+ office employees, accounting firms 10+, law firms 6+.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "The greater of $50,000 or 6% of the licence fees paid to the company for an introducer's first five qualified US referrals, 6% outside the US, paid after the transaction completes. Mode reviews existing claims before assigning credit; the binding agreement is not published.",
      "introducerTerms": {
        "rate": 0.06,
        "flat": 50000,
        "greaterOf": true
      },
      "record": {
        "tier": "partial",
        "says": "Its parent files with the SEC and reported $26.5 million of revenue in the first half of 2026 from its other business. Buying company records began on 13 Oct 2026 with no deal yet, and the same filing discloses doubt about it continuing as a going concern, which management says is alleviated.",
        "source": "https://www.sec.gov/Archives/edgar/data/1748441/000149315226046559/form1-sa.htm",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://bizbuydata.com/50k",
      "sourceLabel": "BizBuyData, referral and seller page",
      "readOn": "2026-10-15"
    },
    {
      "id": "appen",
      "name": "Appen",
      "domain": "appen.com",
      "kind": "principal",
      "what": "A listed supplier of training data to AI labs that also licenses companies' operational records and handles the extraction, anonymisation and the conversations with the labs.",
      "wants": "Operational documentation (procedures, wikis, QA), decision and escalation records (support desk, ITSM, CRM, project tracking, messaging) and systems of record (ERP, warehouse, contracts, HR, procurement).",
      "pays": "No figure published. \"Terms are agreed in writing before extraction begins\"; price follows uniqueness, how connected the systems are, volume and how far back the records go.",
      "low": null,
      "high": null,
      "paidWhen": "\"Payment is on acceptance rather than submission.\"",
      "scrub": "Appen covers extraction, anonymisation and transfer at no cost to the company; personal data is removed, with synthetic rewrite where the workflow has to survive the redaction. The company approves a sample pack before anything is licensed.",
      "criteria": {},
      "eligibility": "Companies of any size; those winding down are answered first.",
      "takes": [
        "company-records"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "Listed on the ASX (APX) and reported $119.9 million of operating revenue for the first half of 2026. No company has said in public that it sold records to Appen's enterprise programme.",
        "source": "https://www.asx.com.au/asxpdf/20260827/pdf/0739ct3d5q4g32.pdf",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.appen.com/enterprise-data",
      "sourceLabel": "Appen, enterprise data page",
      "readOn": "2026-10-15"
    },
    {
      "id": "centific",
      "name": "Centific",
      "domain": "centific.com",
      "kind": "principal",
      "what": "A supplier of training data and expert work to AI labs that licenses companies' operational records for frontier labs' training and pays the company for them.",
      "wants": "Connected records from finance and accounting, healthcare operations (claims, care coordination, de-identified clinical documentation), sales and deal workflows, people operations and IT service management.",
      "pays": "No figure published. Price follows volume, the applications connected, the complexity of the work, data quality and whether comparable data exists.",
      "low": null,
      "high": null,
      "paidWhen": "\"We agree to the terms in writing before extraction begins. Payment is made when the data is accepted.\"",
      "scrub": "Centific pays for extraction, anonymisation and transfer, with no setup fees or deductions; extraction is automated and no person reads the raw content. The company reviews an anonymised sample first.",
      "criteria": {},
      "eligibility": "Enterprises in any industry. Companies winding down get a faster process under an asset-purchase agreement and may license their whole history.",
      "takes": [
        "company-records"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "Raised a $60 million Series A led by Granite Asia (Built In Seattle, 25 Jun 2025). No sale of company records to it has been confirmed by a seller.",
        "source": "https://www.builtinseattle.com/articles/centific-raises-60m-series-a-20250625",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.centific.com/enterprise-data",
      "sourceLabel": "Centific, enterprise data page",
      "readOn": "2026-10-15"
    },
    {
      "id": "collinear",
      "name": "Collinear AI",
      "domain": "collinear.ai",
      "kind": "principal",
      "what": "A lab that builds simulated workplaces and RL environments for AI agents and licenses companies' code and business-tool history to build them.",
      "wants": "Private repositories with pull requests and linked tickets, CI logs and coding-agent sessions; history from Slack, Gmail, Drive, Notion, HubSpot, Salesforce, Linear, Jira and similar tools.",
      "pays": "No figure published; \"We usually quote an initial range within a day.\" The price rises with 15 or more people at peak, several years of history, pull requests with their tickets, human-written code and the same work visible across several tools.",
      "low": null,
      "high": null,
      "paidWhen": "\"We confirm delivery is complete, then pay according to the contract.\" About a week from first call to signature.",
      "scrub": "Collinear removes personal data after export: names become consistent pseudonyms and identifiers are stripped. A one-time export; ownership stays with the company.",
      "criteria": {
        "code": {
          "humanWritten": true
        }
      },
      "eligibility": "Companies that hold the rights to the data, operating ones included (the export is a one-time snapshot). It does not want AI-generated code, a public framework with a thin layer on top, or loose documents with no company around them.",
      "takes": [
        "source-code",
        "company-records"
      ],
      "introducerFee": null,
      "record": {
        "tier": "own-claims",
        "says": "Its site names no customer, and we found no funding round or deal reported by anyone else.",
        "source": "https://collinear.ai/about-us",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://collinear.ai/data-acquisition",
      "sourceLabel": "Collinear AI, data licensing page",
      "readOn": "2026-10-15"
    },
    {
      "id": "labelbox",
      "name": "Labelbox (Prism)",
      "domain": "labelbox.com",
      "kind": "principal",
      "what": "Labelbox's Prism acquires real operational data and work documents from companies and contributors for frontier AI training.",
      "wants": "Work documents (coding sessions, scientific and financial workflows, research traces) and operational systems (Salesforce, Jira, Slack, SAP, ServiceNow, Zendesk, CAD, financial workbooks). Its page marks financial and scientific reasoning traces and cross-system workflow data as urgent.",
      "pays": "\"Smaller targeted datasets starting around $50K and large enterprise partnerships reaching $1M+\".",
      "low": 50000,
      "high": 1000000,
      "paidWhen": "\"From first call to first wire in under 21 days\", with a written quote within 48 hours of scoping.",
      "scrub": "Prism handles ingestion, normalisation and PII scrubbing; data is filtered, anonymised, aggregated or permissioned before use.",
      "criteria": {},
      "eligibility": "Companies in any industry; its page names software, financial services, healthcare, manufacturing and logistics as where demand is highest. Licences are exclusive by default.",
      "takes": [
        "company-records",
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "confirmed",
        "says": "A Series E of more than $100 million in equity and a credit facility at a $1.25 billion pre-money valuation, led by Infinitum Partners (Labelbox, 14 Oct 2026), after a $110 million Series D led by SoftBank Vision Fund 2 (2022).",
        "source": "https://labelbox.com/blog/labelbox-series-e",
        "readOn": "2026-10-16"
      },
      "basis": "stated",
      "source": "https://prism.labelbox.com/",
      "sourceLabel": "Labelbox Prism, home page",
      "readOn": "2026-10-15"
    },
    {
      "id": "general-data",
      "name": "The General Data Company",
      "domain": "thegeneraldata.com",
      "kind": "principal",
      "what": "A San Francisco research lab (General Data Labs, Inc.) that licenses industrial companies' operating records and builds training data and RL environments from them for AI labs and manufacturers.",
      "wants": "Company archives (projects, correspondence, orders, technical documentation), operational processes (complaints, inspection and maintenance records, tickets, order and stock exports) and expertise (drawings, CAD models, change requests, commissioning records).",
      "pays": "No figure published; scope, uses, recipients and payment are set out in writing.",
      "low": null,
      "high": null,
      "paidWhen": "\"Payment follows the agreed milestones.\" No file moves before a signed agreement.",
      "scrub": "It replaces names, companies, addresses and identifiers with invented ones using its own model, a person checks the result, then the link to the real names is deleted.",
      "criteria": {
        "industries": [
          "Manufacturing",
          "Industrial engineering"
        ]
      },
      "eligibility": "Industrial companies, including companies that are no longer trading. It does not take current designs, prices or anything a competitor could use.",
      "takes": [
        "company-records"
      ],
      "introducerFee": "Up to $50,000 when a data partnership is signed, under a written referral agreement; the amount depends on the size of the partnership.",
      "introducerTerms": {
        "flat": 50000,
        "upTo": true
      },
      "record": {
        "tier": "own-claims",
        "says": "A website and its own statements only: we found no funding, named team or deal reported by anyone else.",
        "source": "https://thegeneraldata.com/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://thegeneraldata.com/data-partnerships/",
      "sourceLabel": "The General Data Company, data partnerships page",
      "readOn": "2026-10-15"
    },
    {
      "id": "magma",
      "name": "Magma",
      "domain": "magmahq.ai",
      "kind": "marketplace",
      "what": "Packages the interaction data of companies that build AI agents (traces, corrections, completed workflows) and licenses it to AI labs and researchers for a revenue share.",
      "wants": "Agent traces, corrections and completed workflows from a product's own logs, through a read-only connection to the scope the company approves.",
      "pays": "\"A revenue share, paid as your data licenses.\" No percentage published.",
      "low": null,
      "high": null,
      "paidWhen": "As the data licenses, under the data agreement signed during onboarding.",
      "scrub": "The company sets the scope; handling and retention are left to the data agreement. The page does not describe anonymisation.",
      "criteria": {
        "industries": [
          "Companies building AI agents"
        ]
      },
      "eligibility": "Companies building AI agents.",
      "takes": [
        "company-records"
      ],
      "introducerFee": null,
      "record": {
        "tier": "partial",
        "says": "Listed in Y Combinator's Summer 2026 batch with one named founder. No licence of agent data through it has been reported.",
        "source": "https://www.ycombinator.com/companies/magma",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://magmahq.ai/",
      "sourceLabel": "Magma, home page",
      "readOn": "2026-10-15"
    },
    {
      "id": "opendatabay",
      "name": "Opendatabay",
      "domain": "opendatabay.com",
      "kind": "marketplace",
      "what": "A UK data marketplace where providers list datasets, including AI training data, for self-serve sale, or have Opendatabay broker them.",
      "wants": "Datasets of any kind: its marketplace lists AI and LLM training data, robotics data, synthetic and premium data.",
      "pays": "The seller keeps 70% of a sale up to $100, rising to 95% above $250,000 (the commission falls from 30% to 5% as the sale grows). In a brokered sale the seller gets the agreed provider price and Opendatabay sets its own price to the buyer.",
      "low": null,
      "high": null,
      "paidWhen": "Within 30 days of the end of the month of the sale, once the balance reaches $50.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Data providers with a verified business account. Commission is also owed on sales it introduced that close off the platform.",
      "takes": [
        "datasets-environments",
        "text-archives",
        "images",
        "video",
        "audio-speech"
      ],
      "introducerFee": null,
      "record": {
        "tier": "own-claims",
        "says": "Registered in England and Wales (company 15711573). We found no sale through it, funding or press reported by anyone else.",
        "source": "https://www.opendatabay.com/legal/fees",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.opendatabay.com/legal/fees",
      "sourceLabel": "Opendatabay, fees and commission schedule",
      "readOn": "2026-10-15"
    },
    {
      "id": "pre-dev",
      "name": "pre.dev",
      "domain": "pre.dev",
      "kind": "marketplace",
      "what": "A coding-agent company whose lab sells RL coding tasks built from licensed private code to frontier labs, and licenses private repositories from developers and companies.",
      "wants": "Private codebases with their git history, chosen repository by repository from a connected Git account.",
      "pays": "Its earnings example prices source code at $100 per million tokens and git history at $40 per million tokens, $6,000 for the example repository; each repository gets its own value and grade.",
      "low": null,
      "high": null,
      "paidWhen": "\"Paid Net 15 after funds land\", under one agreement signed once.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Individuals, agencies and companies with codebases they own. Non-exclusive by default.",
      "takes": [
        "source-code"
      ],
      "introducerFee": null,
      "record": {
        "tier": "own-claims",
        "says": "Its own figure of 10,000+ licensed private codebases is not confirmed by anyone else, and we found no funding reported.",
        "source": "https://pre.dev/",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://pre.dev/#code-licensing",
      "sourceLabel": "pre.dev, code licensing section",
      "readOn": "2026-10-15"
    },
    {
      "id": "versos",
      "name": "Versos AI",
      "domain": "versos.ai",
      "kind": "marketplace",
      "what": "A platform that indexes video libraries in place and licenses datasets cut from them to AI labs, keeping track of where each asset goes.",
      "wants": "Large video libraries in cloud storage; its estimator says scripted, multi-camera and first-person footage commands the highest rates and generic stock is commoditising.",
      "pays": "No figure published; its estimator values one non-exclusive licence of the footage that passes buyer filters, at what it calls current market rates.",
      "low": null,
      "high": null,
      "paidWhen": "Not published.",
      "scrub": null,
      "criteria": {},
      "eligibility": "Video libraries of 5,000 hours or more; content stays in the owner's storage.",
      "takes": [
        "video"
      ],
      "introducerFee": null,
      "record": {
        "tier": "partial",
        "says": "Raised a $1.85 million seed led by Innovobot with the New Brunswick Innovation Foundation (BetaKit, Dec 2025). It says it works with frontier labs but names none.",
        "source": "https://www.nbif.ca/nbif-news/versos-ai-announcement",
        "readOn": "2026-10-15"
      },
      "basis": "stated",
      "source": "https://www.versos.ai/",
      "sourceLabel": "Versos AI, home page",
      "readOn": "2026-10-15"
    }
  ]
}