
Fourteen counterparties are on record as licensing content to Meta from 2025-12-05 to 2026-07-27, and none has disclosed an amount.
Every Meta figure here is a dated statement by Meta, a counterparty, a court or a regulator, checked 2026-11-09. This independent reference on AI training data licensing covers Meta.
Each number carries its source and check date, and how DataPacts sources and dates every figure is set out in the methodology.
What Meta Acquires for AI Training: Licenses, Crawling, Users and Human Data
Meta names no data-buying program for outside holders; its routes on record are content licenses, its own crawler, its users' public content and human-made training data.
No seller program, payout or referral fee for Meta is recorded in the sources checked on 2026-11-09. The relationships on record are content agreements with named publishers and access to Wikimedia Enterprise's paid API, described below.
- Licensed content: Meta's post of 2025-12-05 names news partners for Meta AI answers, and Wikimedia Enterprise named Meta a paid API customer on 2026-01-15.
- Crawling: Meta's documentation says Meta-ExternalAgent collects content for training foundation AI models or indexing.
- Own users: Meta announced on 2025-04-14 that it would train on public content of adults in the EU, with an objection form.
- Human data and investment: Meta's Llama 2 paper counts human annotations, and Scale AI announced on 2025-06-12 that Meta holds a minority stake in Scale AI.
Meta's summary of training content under Article 53 of the EU AI Act is not located, and the European Commission's list of General-Purpose AI Code of Practice signatories, 22 names updated 2026-10-07, does not mention Meta (checked 2026-11-09).
Meta Company Identity
Meta's US parent is Meta Platforms, Inc., a Delaware corporation registered with the SEC under CIK 0001326801, and Meta Platforms Ireland Limited contracts with businesses outside the US and Canada under the WhatsApp Business Platform terms (s. 1.2). The identity table lists each attribute with its source, as of 2026-11-09.
| Attribute | Value | Source |
|---|---|---|
| Legal name | Meta Platforms, Inc.; EU entity Meta Platforms Ireland Limited | SEC EDGAR |
| Registry | SEC CIK 0001326801, Delaware | SEC EDGAR |
| Website | https://about.meta.com | about.meta.com |
| Headquarters country | United States | SEC EDGAR |
| Founded | Not recorded | n/a |
| Other entities in a regulator record | Facebook UK Limited and Facebook Ireland Limited, named with Meta Platforms, Inc. in the CMA case | gov.uk |
| Type | Frontier lab; developer of Llama and Meta AI; minority investor in Scale AI | about.meta.com |
The founding date is not recorded here because no filing or history page was checked for it.
Meta Program Terms, Payouts and Referral Fees, Dated
Meta publishes no data partnership program, no payout and no referral fee; the only terms Meta states for outside content are those of its Meta AI news partners and its crawler documentation. The program terms table shows each term with what was found, as of 2026-11-09.
| Term | What Meta states | Checked | Source |
|---|---|---|---|
| Seller program for outside data holders | None found | 2026-11-09 | about.meta.com |
| Payout or price | None published | 2026-11-09 | n/a |
| Referral fee | None recorded | 2026-11-09 | n/a |
| News partner scope | Information and links in Meta AI answers; training and financial terms not mentioned | 2026-11-09 | about.fb.com |
| License term, exclusivity, deletion | Not stated | 2026-11-09 | n/a |
| Crawler use | Training foundation AI models or indexing content; blocked by a robots.txt disallow rule | 2026-11-09 | developers.facebook.com |
| User-initiated fetcher | Meta-ExternalFetcher may bypass robots.txt rules | 2026-11-09 | developers.facebook.com |
A licensor cannot learn from Meta's pages what a license would pay, what it would cover, how long it would run, whether it would be exclusive, whether content must be deleted afterwards or when payment would fall due. No source checked states any of these terms for the agreements below.
Other buyers publish more. Dated payout ranges, referral fees and license scope for the buyers that state them are set out in the program terms compared, where Meta has no row of its own to fill.
Who Licenses Content to Meta: 14 Agreements on Record
Fourteen counterparties are on record as licensing content to Meta: thirteen news publishers that Meta names as Meta AI partners between 2025-12-05 and 2026-07-27, and Wikimedia Enterprise, which named Meta a paid API customer on 2026-01-15. Five publishers were named in the first wave of 2025-12-05, and eight more appear in Meta's post as updated on 2026-07-27.
The table lists the agreements newest first, then alphabetically within a date, as of 2026-11-09. The use column reads D for answers or display and n/s for not stated; "listed" means first named in Meta's update of 2026-07-27, because no signing date is published.
| Date | Licensor | What is licensed | Use | Value | Source |
|---|---|---|---|---|---|
| 2026-07-27 (listed) | Le Figaro | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | News Corp | News content for Meta AI; CEO, 2026-11-04: "We are pleased to have partners in OpenAI and Meta" | D | Not disclosed by the parties | about.fb.com; sec.gov |
| 2026-07-27 (listed) | Newsmax | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | PRISA | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | Süddeutsche Zeitung | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | The Daily Caller | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | The Washington Examiner | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-07-27 (listed) | Fox Sports | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2026-01-15 | Wikimedia, LLC (Wikimedia Enterprise) | Wikipedia and Wikimedia content via the paid Enterprise API (on-demand, snapshot, realtime) | n/s | Not disclosed | enterprise.wikimedia.com |
| 2025-12-05 | CNN | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2025-12-05 | Fox News | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2025-12-05 | Le Monde Group | News in Meta AI answers | D | Not disclosed | about.fb.com |
| 2025-12-05 (People Inc.: "signed in Q4 2025") | People Inc. | Publisher content for Meta AI; part of a 49 % rise in licensing and other revenue in Q3 2026 | D | Not disclosed | about.fb.com; sec.gov |
| 2025-12-05 | USA TODAY Co. (USA TODAY Network listed 2026-07-27) | Real-time news in Meta AI answers | D | Not disclosed | about.fb.com |
No agreement in the table carries a value stated by Meta or the licensor. People Inc. attributes part of a 49 % rise in licensing and other revenue in Q3 2026 to its Meta content partnership, without an amount, and News Corp's chief executive names Meta among his company's AI partners without a price.
The tracker of AI data licensing deals records agreements across buyers with the same rule that only a stated value is shown.
What Meta's Content Agreements Permit: Answers and Links, Training Not Stated
Meta describes its news agreements as bringing real-time information and links into Meta AI answers, and its post mentions neither training nor payment terms. Three points follow from the record.
- Meta's post of 2025-12-05, as updated 2026-07-27, names the partners and the use in answers and links.
- The post states no license term, price, exclusivity or training right, so none is attributed to the agreements.
- Wikimedia Enterprise describes a paid API with on-demand, snapshot and realtime access, and states no pricing.
A holder weighing a comparable arrangement would write the permitted use, term, exclusivity and deletion into the AI data licensing agreement itself, because none of those terms can be read from Meta's public statements.
Meta's Own Users as a Training Source: Public Posts and the EU Objection
Meta announced on 2025-04-14 that it would train AI on public content shared by adults in the EU and on their interactions with Meta AI, subject to an objection form.
The Irish Data Protection Commission stated on 2025-05-21 that the training was scheduled to begin on 2025-05-27, as reported by the DPC, and the actual start is not confirmed.
The Higher Regional Court of Cologne rejected an interim application by Verbraucherzentrale NRW on 2025-05-23 (OLG Köln 15 UKl 2/25), and main proceedings were not found as of 2026-11-09. noyb states that it filed 11 complaints with data protection authorities on 2024-06-06.
Whether an individual or a company can stop the use of public posts, and how, is covered under opting out of AI training and its limits. Company data that is not public content on Meta's services is not part of this route.
Content Meta Collects Without a License: Crawlers and Commercial Terms
Meta states that Meta-ExternalAgent crawls "for training foundation AI models or improving products by indexing content directly" and that a robots.txt disallow rule blocks it (checked 2026-11-09).
Meta-ExternalFetcher is a separate agent that fetches pages when a user asks, and Meta states that it may bypass robots.txt rules. Cloudflare measured Meta-ExternalAgent at 7.7 % of crawler traffic to its North American customers in October 2025 data, which is Cloudflare's measurement and not Meta's.
The UK Competition and Markets Authority opened an investigation on 2026-10-30 under Chapter II of the Competition Act 1998 into whether terms that let Meta collect content from business and publisher websites for products including AI features are exploitative. The investigation is open, and no outcome is recorded.
Meta Reputation: Lawsuits and Regulators Over How It Acquired Data
Six proceedings with a primary source, listed here, concern how Meta acquired training data: three US suits over books and films and three European, Brazilian or UK proceedings, listed with their status as of 2026-11-09. The table gives the forum, the docket where one exists, the filing date and what is alleged or decided.
| Proceeding | Forum and docket | Filed or opened | What is alleged or decided | Status as of 2026-11-09 | Source |
|---|---|---|---|---|---|
| Kadrey et al. v. Meta Platforms, Inc. | N.D. Cal. 3:23-cv-03417-VC | 2023-07-07 | 13 authors; training Llama on books from shadow libraries; Dkt. 598 (2025-06-25) summary judgment for Meta on training on that record | Related cases coordinated; Entrepreneur Media trial 2027-05-03 (Dkt. 879) | courtlistener.com |
| Elsevier, Cengage, Hachette, Macmillan, McGraw Hill and others v. Meta Platforms, Inc. and others | N.D. Cal. 3:26-cv-11415-VC (transferred from S.D.N.Y. 2026-09-25) | 2026-05-05 | Alleged copyright infringement in training Llama | Pending, no trial date | plaintiffs' counsel |
| Strike 3 Holdings, LLC and Counterlife Media, LLC v. Meta Platforms, Inc. | N.D. Cal. 5:25-cv-06213-EKL | 2025-07-23 | Alleged BitTorrent downloading of at least 2,396 films to train Movie Gen, LLaMA and other models | Pending | complaint |
| Verbraucherzentrale NRW v. Meta | OLG Köln 15 UKl 2/25 | n/r | Interim injunction against training on Facebook and Instagram data rejected 2025-05-23 | Main proceedings not found | nrwe.justiz.nrw.de |
| ANPD v. Meta | ANPD, Voto nº 11/2024/DIR-MW/CD | 2024-07-02 | Preventive suspension of processing for AI training, daily fine BRL 50,000 | Later status not recorded | gov.br/anpd |
| CMA investigation of Meta's commercial terms | CMA, Competition Act 1998 Chapter II; no reference published | 2026-10-30 | Whether terms letting Meta collect publisher and business website content for AI features are exploitative | Open | gov.uk |
The Kadrey docket and its related cases are the closest to a ruling on the training question, and the tracker of AI copyright lawsuits over training data places them beside the other US decisions; a profile of Kadrey v. Meta follows the claims and rulings in full.
For licensors, the Kadrey order of 2025-06-25 contains two statements that point in different directions.
The court wrote that the plaintiffs were "not entitled to the market for licensing their works as AI training data", and also that the ruling "does not stand for the proposition that Meta's use ... is lawful". Both sentences concern the record before that court.
Later steps in several of these proceedings are not recorded here, and the page states only what the listed sources show. The ANPD order is recorded without any later decision.
Is Meta a Realistic Licensee for Your Data? Benefits, Risks, Downsides and Alternatives
Meta licenses content from large publishers and an encyclopedia API, but it runs no door for other holders, publishes no terms and has disclosed no price.
Benefits
- Meta has named 13 news publishers between 2025-12-05 and 2026-07-27, and Wikimedia Enterprise names Meta a paid API customer.
- A counterparty can state the revenue effect, as People Inc. does for Q3 2026.
- Named agreements show that a large lab accepts display and answer use as a licensed purpose.
Risks
- Court and regulator proceedings over how Meta acquired books, films and personal data are open (Kadrey, 3:26-cv-11415-VC, 5:25-cv-06213-EKL, the CMA case).
- No termination, deletion or audit terms are public.
- Business-chat data from the WhatsApp Business Platform cannot be licensed under Meta's own terms (see the supplementary section).
Downsides
- No seller program or application route is found, so a holder has no stated way to approach Meta.
- The agreements on record state answers and links, and no price or payout is public.
- Meta's post does not mention training, so a holder cannot assume that a Meta license covers it.
Alternatives
- Other labs that run or publish programs, profiled under the labs hub.
- Blocking Meta-ExternalAgent with a robots.txt disallow rule, which Meta states blocks that agent.
- Not licensing at all, which leaves the data with the holder and is a valid outcome.
Holders who prefer not to approach a lab can compare alternatives to selling data, including keeping it, where not selling is treated as an option on equal terms.
Meta Among Frontier AI Labs That Acquire Data
Meta is one of 20 AI developers profiled as data buyers on DataPacts, and its recorded content agreements name each counterparty, unlike its crawling. The hub on frontier AI labs as data buyers compares all 20, and the lab profiles for OpenAI, Google, Anthropic and Microsoft show how other labs state their terms.
The full directory answers who buys AI training data, and the page on news publishers licensing content to AI companies covers the seller side. A holder deciding between buyers can use the guidance on choosing a buyer for your data.
- Meta AI news licensing partners: the program-level record of the publishers above
- Meta-ExternalAgent and Meta's other bots, with example rules in robots.txt for AI crawlers
- Kadrey v. Meta in full, and the Meta AI training on EU user data proceedings
More About Meta as a Data Buyer
Meta and Human Data: Llama 2 Annotations and Vendor Claims
Meta's Llama 2 paper of 2023-07-19 counts 27,540 supervised fine-tuning annotations and 1,418,091 human preference comparisons Meta collected, 2,919,326 including open-source sets. Gray Swan states, in its announcement of 2026-05-28, that Meta is among its users. Vendors that sell human data to labs are compared on their own pages.
Business Messages on WhatsApp: What Meta's Platform Terms Forbid
A business that holds customer chats from the WhatsApp Business Platform cannot license them for AI training: Meta's terms (last modified 2026-09-23) prohibit selling or licensing Platform Data, including anonymous, aggregate and derived forms (s. 4.1(b)), and using it to train AI except fine-tuning for the business's exclusive use (s. 4.7).
The page on WhatsApp Business chats as AI training data sets out the terms in full, and this is information, not legal advice.
Frequently Asked Questions About Meta and Data Licensing
Does Meta pay for training data?
Wikimedia Enterprise names Meta a customer of its paid API and People Inc. reports licensing revenue from its Meta partnership, but no party has disclosed an amount and Meta's post does not mention training.
Can a company license data to Meta?
No seller program or application route for outside data holders was found in the sources checked on 2026-11-09. Meta's recorded agreements are with news publishers and Wikimedia Enterprise.
Did Meta buy Scale AI?
No: Scale AI announced on 2025-06-12 that Meta would hold a minority of its outstanding equity, valuing Scale at over USD 29 billion. Scale states no percentage.
Are Meta's news agreements training deals?
Meta describes them as bringing real-time information and links into Meta AI answers. Its post mentions neither training nor payment terms.