Broker directory
The middlemen between data holders and AI labs, grouped by how they make money, which tells you whose side they're on. BlobCap isn't one of them: we don't buy, sell or take a cut.
Brokers 40
- TroveoMarketplacetroveo.ai ↗
Licensing marketplace for real-world video and audio from creators and media companies, now also brokering company data.
How you get paidPaid when your content licenses. Creators report $0.75–3 per minute of footage.
PublicSource ↗ - Calliope NetworksMarketplacecalliopenetworks.ai ↗
Signs YouTube creators to licensing contracts and sublicenses their videos to AI companies.
CreatorsVideoHow you get paidRevenue share: a portion of the royalties from each sublicense.
PublicSource ↗ - WirestockMarketplacewirestock.io ↗
Community of 700,000+ photographers and artists whose work is licensed for AI training.
How you get paidFrom a few cents per item up to $15–20 per artwork for specific requests.
PublicSource ↗ - Human NativeMarketplaceAcquired by Cloudflare, January 2026.humannative.ai ↗
UK marketplace matching rights holders with AI developers who want training and inference data.
How you get paidRights holders set prices and are paid for licensed content.
PublicSource ↗ - ProtegeMarketplacewithprotege.ai ↗
Platform for data partnerships between rights holders and AI labs. Raised $70M.
CompaniesPublishersHow you get paidNot disclosed
PublicSource ↗ - versos aiMarketplaceversos.ai ↗
Marketplace for rights-cleared video training datasets.
CreatorsVideoHow you get paidNot disclosed
ClaimedSource ↗ - Defined.aiMarketplacedefined.ai ↗
Long-running catalog of commissioned and curated datasets, strongest in speech.
How you get paidHas reported buyers paying up to $2 per image and $100–300 per hour of long-form video.
ClaimedSource ↗ - HUD DataVendorMarketplacedatavendor.ai ↗
Marketplace for startup codebases. Grades repos, then sells them non-exclusively to AI companies. Doesn't take email, chat or CRM data yet.
CompaniesCode & repositoriesHow you get paidYou set the price and receive 80% by wire or ACH after delivery.
Fees & termsKeeps 20% of each sale. Doesn't name its buyers.
ClaimedSource ↗ - LabelsetsMarketplace
Marketplace for labeled training datasets across vision, text, medical, financial and audio. Launched April 2026.
CompaniesIndividualsHow you get paidWeekly payouts. You keep 85% of each sale.
Fees & terms15% commission; free to list.
ClaimedSource ↗ - OpendatabayMarketplaceopendatabay.com ↗
Licensed marketplace for AI-ready datasets with due-diligence scoring and provenance checks. Registered businesses only.
CompaniesHow you get paidNot disclosed
ClaimedSource ↗ - DataradeMarketplacedatarade.ai ↗
Large B2B data marketplace connecting data providers with buyers, including AI training data.
CompaniesHow you get paidNot disclosed
Fees & termsOffers a commission-only listing tier.
ClaimedSource ↗ - Created by HumansMarketplacecreatedbyhumans.ai ↗
Lets authors claim their books and license AI training and reference rights. Launched with the Authors Guild.
How you get paid80% of proceeds go to the author.
Fees & termsNon-exclusive contract.
PublicSource ↗ - Avail CorpusMarketplace
Product for smaller media companies and independent creators to license their content to AI companies.
How you get paidNot disclosed
PublicSource ↗ - SourceAudioMarketplacesourceaudio.com ↗
Opt-in AI dataset marketplace for 14M cleared tracks and 3M sound effects. Rightsholders see each deal's terms before opting in.
How you get paidPaid up front or over the license term, per deal.
ClaimedSource ↗ - LANDR Fair Trade AIMarketplacelandr.com ↗
Opt-in program for LANDR Distribution artists to license their tracks for AI training. Artists can opt out anytime.
CreatorsMusicHow you get paid25% of net licensing revenue, split by contribution, plus a recoupable $5-per-track advance.
Fees & termsOnly tracks where you keep your publishing rights qualify.
PublicSource ↗ - Shutterstock Contributor FundMarketplaceshutterstock.com ↗
Pays Shutterstock contributors when their work is included in datasets licensed to AI companies.
How you get paidPaid every six months, based on how much of your work was used.
PublicSource ↗ - micro1Buys and resellsmicro1.ai ↗
Pays companies for their workplace data and supplies it to AI labs. Says it plans to spend $1B on company data in 12 months.
How you get paidOne-time payment, reported at $100K–$2M per company.
Fees & termsOffers often come with 48-hour deadlines.
PublicSource ↗ - MercorBuys and resellsmercor.com ↗
AI data vendor that hires experts and also buys company archives. Bid on Spirit Airlines' data and lost to Google.
How you get paidHas offered professionals one-time payments of up to $10K for workplace chat history.
Fees & termsReported to be buying old work materials that often belong to employers.
PublicSource ↗ - Kled AIBuys and resellskled.ai ↗
App where people upload their files. Kled pays them, then packages the data into datasets for AI companies.
How you get paidPaid per upload in the app.
PublicSource ↗ - AnthrolyBuys and resellsanthroly.com ↗
Licenses operating data directly from the companies that produce it and pays for it itself.
How you get paidPays the company directly. Says it has transacted $900K+ in data.
ClaimedSource ↗ - Surge AIBuys and resellssurgehq.ai ↗
Major AI data vendor that buys wound-down startups' workspace data to build RL environments for labs.
How you get paidNot disclosed
PublicSource ↗ - Handshake AIBuys and resellsjoinhandshake.com/ai/data-partnerships ↗
Licenses de-identified company data to frontier labs. Looks for companies with 20+ full-time employees and 3+ years in operation.
How you get paidOne-time payment, $100K–$4M+ depending on the data.
Fees & termsPays $10K for introductions that lead to a partnership.
ClaimedSource ↗ - Turing (Project Lazarus)Buys and resellslazarus.turing.com ↗
Buys codebases and operating records (design docs, test logs, wikis) from failed and active startups for frontier model training.
How you get paidSingle payout of up to $1M. Founders set the approved scope.
ClaimedSource ↗ - Ooak DataBuys and resellsooakdata.com ↗
Y Combinator startup that licenses business data, anonymizes it and builds agent training environments from it.
How you get paidUp to $300K. Closed deals so far range from $20K to several hundred thousand.
ClaimedSource ↗ - NyneBuys and resellsnyne.ai ↗
Evaluates historical company data and licenses it to AI labs. Focuses on companies with 20+ US W-2 employees and 3+ years in business.
How you get paidNot disclosed
ClaimedSource ↗ - Luel AIBuys and resellsluel.ai ↗
App that pays people for voice and video recordings, then licenses them as training data. Raised $31M.
How you get paidPaid per recording in the app.
PublicSource ↗ - AIData.BrokerDeal brokeraidata.broker ↗
Values your data, checks rights, prepares a datasheet and presents it to buyers you approve. Also publishes price medians.
CompaniesCreatorsPublishersCompany workspace dataCode & repositoriesBusiness recordsBooksImagesVideoSpeech & audioHow you get paidThe buyer pays you; the broker takes its fee from the sale.
Fees & terms15% of the first $250K, less above that. Only charged if a sale closes.
ClaimedSource ↗ - PolysharesDeal broker
Licenses business operating records to AI labs. Doesn't tell sellers which lab buys which data.
How you get paidNot disclosed
Fees & termsHas pitched offers at "double market" with short deadlines.
PublicSource ↗ - FileYieldDeal brokerfileyield.com ↗
Evaluates your data, strips personal information, then runs competing bids from buyers and negotiates terms for you.
CompaniesCreatorsHow you get paidNot disclosed
ClaimedSource ↗ - SimpleClosure Asset HubWind-down servicesimpleclosure.com ↗
Startup shutdown service that added code and data sales in April 2026. 150+ deals; 47% of its wind-downs now include a data sale.
How you get paidProceeds usually go to creditors and investors first. Reported payouts $10K–100K.
PublicSource ↗ - SunsetWind-down service
Wind-down service with a fast-growing data resale business.
How you get paidValues assets by headcount, years in operation, data volume and how the team used its tools.
PublicSource ↗ - TollBitLicensing platformtollbit.com ↗
Access layer where AI companies pay publishers per page they fetch.
PublishersNews & publishingHow you get paidPublishers set per-page prices for summarization or full-display use. Not a training license.
PublicSource ↗ - ProRataLicensing platformprorata.ai ↗
Pays publishers based on how much of their content shows up in AI-generated answers.
PublishersNews & publishingHow you get paidShare of ad revenue from AI answers that cite your content.
PublicSource ↗ - Cloudflare Pay Per CrawlLicensing platformcloudflare.com ↗
Lets site owners charge AI crawlers per request instead of blocking them. Private beta since July 2025.
How you get paidPer crawl, at prices the site owner sets.
PublicSource ↗ - PLAI by GEMALicensing platformgema.de/en/about-gema/ki-hub/plai-by-gema ↗
German music rights society's licensed training dataset, with composition and recording rights bundled. About 178,000 tracks across 60 genres.
PublishersMusicHow you get paidThrough participating publishers.
PublicSource ↗ - Microsoft Publisher Content MarketplaceLicensing platformmicrosoft.com ↗
Marketplace where publishers are paid when AI products use their content, starting with Copilot as the first buyer.
PublishersNews & publishingHow you get paidNot disclosed
PublicSource ↗ - VerbPersonal data exchange
Phone app that lets people sell activity data they choose to share. Expects AI labs to be its biggest buyers.
IndividualsHow you get paidYou set the price; Verb suggests one based on what you share.
PublicSource ↗ - VanaPersonal data exchangevana.org ↗
MIT spinout where people pool personal data in "data DAOs" and get rewarded when it's used. 1M+ contributors.
IndividualsHow you get paidTypical payouts have been a few dollars per person.
PublicSource ↗ - Neon MobilePersonal data exchangeneonmobile.com ↗
App that pays people to record their phone calls and resells the audio to AI developers.
IndividualsSpeech & audioHow you get paid15¢ per minute, or 30¢ when both callers use Neon. Capped at $30 a day.
Fees & termsTerms include an arbitration clause and class-action ban.
PublicSource ↗ - SilencioPersonal data exchangesilencio.network ↗
App that collects noise levels and voice recordings from phones and sells the data.
IndividualsSpeech & audioHow you get paidPaid mostly in its own SLC crypto token; some USDC when a client buys voice data.
ClaimedSource ↗
Claimed means the details come from the broker's own site or pitch. Know a broker we're missing, or dealt with one? Add your deal anonymously.