DataPacts

Wikimedia Enterprise AI Deals (Amazon, Meta, Microsoft, Mistral, Perplexity, Google): Terms and Sources

Checked

At a glance

Wikimedia Enterprise, the paid API service of Wikimedia, LLC, a subsidiary of the Wikimedia Foundation, names eleven companies as partners in its post of 2026-01-15, and none of the parties has disclosed a value.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

Google has the earliest dated mention: the Wikimedia Foundation announced on 2022-06-21 that Google and the Internet Archive were Wikimedia Enterprise's first customers, in a post that names no AI or training use by Google.

The Wikimedia Foundation published its own partner list on the same day, and the two lists sort the companies into new and existing differently.

This page reads each Wikimedia post with its date, shows the access tiers and the Wikipedia license terms, and records one openly licensed source that charges for access and service, as one example in AI training data licensing.

Every date and quotation below comes from a Wikimedia post, page or terms document, and each is checked as described in how we source, date and check every figure.

Wikimedia Enterprise's AI partners at a glance: names, first dated mention and source

Eleven companies appear in Wikimedia's own posts, and Google has the earliest dated mention (2022-06-21, as first customer). The table lists them alphabetically as of 2026-11-17, with the source of each row.

Partner First dated mention in a Wikimedia post Status in the Enterprise post of 2026-01-15 Status in the Foundation post of 2026-01-15 Use as stated Value Source
Amazon 2026-01-15 announced for the first time existing not stated not disclosed enterprise.wikimedia.com; wikimediafoundation.org
Ecosia 2026-01-15 (own post 2026-02-16) roster new search results and AI answers (post title) not disclosed enterprise.wikimedia.com
Google 2022-06-21 (first customer) roster existing knowledge panels and other features; AI not stated not disclosed wikimediafoundation.org
Meta 2026-01-15 announced for the first time existing not stated not disclosed enterprise.wikimedia.com; wikimediafoundation.org
Microsoft 2026-01-15 announced for the first time new not stated not disclosed enterprise.wikimedia.com; wikimediafoundation.org
Mistral AI 2026-01-15 (own post 2026-01-27) announced for the first time new "a source of information" for its models; Le Chat named not disclosed enterprise.wikimedia.com
Nomic 2025-06-19 (own post) roster not named visual maps of multilingual Wikipedia (post title) not disclosed enterprise.wikimedia.com
Perplexity 2026-01-15 announced for the first time new not stated not disclosed enterprise.wikimedia.com; wikimediafoundation.org
Pleias 2025-12-15 (review) roster new "to drive open, ethical AI training" not disclosed enterprise.wikimedia.com
ProRata 2025-03-31 (own post) roster new search; "healthier digital content economy" (review) not disclosed enterprise.wikimedia.com
Reef Media 2025-11-05 (own post) roster not named fact checking with the Snapshot API (post title) not disclosed enterprise.wikimedia.com

The date in the second column is the earliest Wikimedia post read for that company, not the date of a contract. No post states a contract start, term, exclusivity or renewal for any company, so the page does not give one.

The 2026-01-15 announcements: two Wikimedia posts, two partner lists

Wikimedia Enterprise and the Wikimedia Foundation published partner lists on the same day, 2026-01-15, and they sort the companies into new and existing differently. The table sets the two wordings side by side as of 2026-11-17.

Post Date Wording First announced or new Existing or roster Source
Wikimedia Enterprise blog 2026-01-15 "announcing ... for the first time as they join our roster of partners, which includes" Amazon, Meta, Microsoft, Mistral AI, Perplexity Google, Ecosia, Nomic, Pleias, ProRata, Reef Media enterprise.wikimedia.com
Wikimedia Foundation news 2026-01-15 "became new Wikimedia Enterprise partners, joining existing partners such as" Ecosia, Microsoft, Mistral AI, Perplexity, Pleias, ProRata Amazon, Google, Meta wikimediafoundation.org

The Enterprise post announces Amazon, Meta, Microsoft, Mistral AI and Perplexity "for the first time as they join our roster of partners", and adds that "over the past year, we have formalized relationships with several companies that rely on Wikimedia project data".

Nomic and Reef Media appear in this roster and not in the Foundation's post. The Enterprise post also describes Wikipedia as "one of the highest-quality datasets for training Large Language Models", which is a general statement and not a statement about any named company.

The Foundation's post says that Ecosia, Microsoft, Mistral AI, Perplexity, Pleias and ProRata "became new Wikimedia Enterprise partners, joining existing partners such as Amazon, Google, and Meta".

Amazon and Meta are therefore announced for the first time in one post and called existing in the other, while Ecosia, Pleias and ProRata are on the roster in one post and new in the other.

This page keeps both wordings visible. It writes "announced for the first time on 2026-01-15" for the five names of the Enterprise post, and it uses "new" or "existing" only together with the post that says it.

Google and the Internet Archive: Wikimedia Enterprise's first customers in 2022

The Wikimedia Foundation announced on 2022-06-21 that Google and the nonprofit digital library Internet Archive were Wikimedia Enterprise's first customers. The post says Wikimedia content helps power some Google features, "including being one of several data sources that show up in its knowledge panels".

The post adds that Enterprise makes content sourcing more efficient for Google. It states no amount and names no AI or training use by Google; AI training appears only as a general example of a possible customer.

Press items that describe the 2022 arrangement as payment for AI training access are not adopted here, because the announcement does not say it. The wider record of Google's programs and payouts sits on the profile of Google as a data buyer.

On the terms, the Internet Archive receives Wikimedia Enterprise "free of cost" according to the same post. For other customers the post describes "a variable price based on their volume of use" and a free trial of 10,000 on-demand requests and a 30-day Snapshot.

It also says that Wikimedia's projects, public datasets, tools and APIs "will continue to be available for free use to all users".

Mistral AI, Pleias and other partners in Wikimedia Enterprise's own posts

Wikimedia Enterprise's blog dates individual partnerships and uses, and two posts state what the content is for in AI terms: Pleias for "open, ethical AI training" and Mistral AI as "a source of information" for its models.

The table lists nine company posts from the 26 posts on the first blog page, dated ascending as of 2026-11-17, with the titles as published.

Date Company Post title (Wikimedia Enterprise)
2025-03-31 ProRata "Wikimedia Enterprise Partners with ProRata.ai to Champion Sustainable Search Engine Practices"
2025-06-19 Nomic AI "Nomic AI's NOMAD Projection uses Enterprise Datasets to Visually Map Multilingual Wikipedia"
2025-11-05 Reef Media "Reef Media uses Wikimedia Enterprise Snapshot API to Fact Check and Verify Sources"
2026-01-27 Mistral AI "Mistral AI partners with Wikimedia Enterprise to Leverage Open Knowledge for AI"
2026-02-16 Ecosia "Ecosia Enriches Search Results and AI Answers with Wikimedia Enterprise"
2026-03-17 Firecrawl "Firecrawl Replaces Wikipedia Scraping with Wikimedia Enterprise APIs"
2026-03-24 Aligned AI "Aligned AI is developing Ethical AI products for families, with the help of Wikimedia Enterprise"
2026-07-07 GNOMI "GNOMI Partners with Wikimedia Enterprise to Bring Trusted Human Knowledge to Its AI News Agent"
2026-09-22 Mantic "Mantic Uses Wikimedia Enterprise to Power AI Forecasting"

The Mistral AI post of 2026-01-27 says the company uses the Snapshot API to fetch a complete database of Wikimedia projects in several languages every day, delivered as Structured Contents JSON, and that "these snapshots are used as a source of information for Mistral AI's models". It names Le Chat and gives no price.

Whether the data trains models or only informs answers is not stated, and the company's other programs are on the profile of Mistral AI as a data buyer. The review of 2025-12-15 says Pleias "entered a partnership to drive open, ethical AI training", which is the one post that names training.

A post titled "uses" does not show a paid account, because free accounts exist: a post of 2026-06-02 announces that free API accounts get more requests and free Structured Contents Snapshots. Only the first page of the blog was read, so older and newer posts may name further companies.

What partners receive: On-demand, Snapshot and Realtime APIs on free or paid accounts

Wikimedia Enterprise offers its content through three APIs: On-demand for the latest version of one article, Snapshot for downloadable files per language, and Realtime for a stream of updates. The table compares the account tiers as stated by Wikimedia Enterprise, as of 2026-11-17.

Item Free account Paid account Source
On-demand API (latest version of one article) up to 50,000 requests per month higher limits, scoped with sales pricing page
Snapshot API (downloadable file per language) monthly updates; up to 30 requests and 1,500 snapshot chunks per month updates up to daily pricing page; partner post ("updated every hour")
Realtime API (stream of updates) not listed included pricing page
Price free "Egress pricing is bespoke"; contact sales pricing page
2022 terms (for comparison) free trial: 10,000 on-demand requests, 30-day Snapshot "variable price based on their volume of use" wikimediafoundation.org, 2022-06-21

The post of 2026-01-15 describes the Snapshot API as Wikipedia "as a downloadable file for every language, updated every hour", while the pricing page lists updates "up to daily" for paid accounts, and both stand as Wikimedia Enterprise's statements.

The same post claims "more than 65 million articles in over 300 languages". How such files move from a seller to a buyer in general, by API, snapshot or transfer, is covered under delivering licensed data to an AI buyer.

Why no value or partner price appears for any Wikimedia Enterprise deal

Neither Wikimedia nor any named company has stated what a partner pays, so this page shows no amount for any deal. No deal values or partner pricing are disclosed in the posts read, and the pricing page publishes free limits only and states that egress pricing is bespoke.

The only financial statement on record is the 2025 review of 2025-12-15, which calls Wikimedia Enterprise "a self-sustaining, profitable initiative this fiscal year" without a figure.

Headlines that call the arrangements "paid deals" or "AI training deals" come from press items and carry no amount and no quotation from either post of 2026-01-15, so this page does not echo them. A reader looking for revenue or price benchmarks finds none in the Wikimedia record.

Wikipedia's CC BY-SA terms and what paid Enterprise access adds

Wikipedia text is licensed under the Creative Commons Attribution-ShareAlike 4.0 International license and the GNU Free Documentation License, so any reuser can take it without an Enterprise contract if it meets those licenses.

The Wikimedia Terms of Use, in effect since 2023-06-07, say that reusers may comply with either license or both, must attribute the authors by link, page history or author list, and must license modified text under CC BY-SA 4.0 or later.

What paid access adds is volume, speed and service, not a different license. The Wikimedia Foundation's post of 2025-11-10 asks AI developers for "attribution and financial support" and says "most AI developers should properly access Wikipedia's content through the Wikimedia Enterprise platform", which it calls a "paid-for opt-in product".

Free routes exist outside Enterprise, among them database dumps, the MediaWiki Action API and Wikimedia Cloud Services, and no page read states that Enterprise changes the content license.

Creative Commons states that its licenses apply only when copyright permission is required, and the terms of CC licenses for training are covered under dataset licenses for AI training such as CC. This page does not say whether training on CC BY-SA text triggers share-alike or attribution duties. It is information, not legal advice.

What the Wikimedia record shows a holder weighing paid API access

Wikimedia's record shows one openly licensed source charging for access and service rather than for rights. The lists below set out what the record supports as of 2026-11-17.

Benefits

  • A paid tier runs beside free use, and Wikimedia Enterprise calls the initiative "self-sustaining, profitable" in its 2025 review.
  • Named companies across labs and search appear in dated posts, so a holder can see who uses the service.

Risks

  • The content is already free under CC BY-SA, so a buyer can choose dumps instead of a paid tier.
  • No stated price exists to benchmark against.

For a nonprofit that holds openly licensed text, the structure matters: Wikimedia offers the service through a subsidiary LLC, and the model is described under nonprofits licensing data through a subsidiary.

Downsides

  • Partner lists differ between two posts of the same day.
  • Use is not stated for most named companies.

Alternatives

  • Not licensing at all.
  • Open release under a public license.
  • Access by free account limits only.
  • Licensing exclusive or non-public material, as in the deals on the tracker.

Whether any of these fits a particular holder depends on its rights and its aims, which the guide on should you sell or license your data to AI companies weighs for each option.

Where to read the buyers, the tracker and other licensed sources around Wikimedia Enterprise

This page covers Wikimedia Enterprise's named partners; the pages below cover each named buyer, every recorded deal and other licensed or open text sources.

Frequently asked questions

Does Google pay Wikipedia for AI training?

The Wikimedia Foundation announced Google as a first customer of Wikimedia Enterprise on 2022-06-21, and the announcement states no amount and no AI training use. It names knowledge panels as one Google use. Press claims of payment for AI training are not in that post.

Which AI companies have Wikimedia Enterprise access?

Wikimedia Enterprise's post of 2026-01-15 names Amazon, Meta, Microsoft, Mistral AI and Perplexity for the first time and lists Google, Ecosia, Nomic, Pleias, ProRata and Reef Media on its roster. The Wikimedia Foundation's post of the same day calls Ecosia, Microsoft, Mistral AI, Perplexity, Pleias and ProRata new, and Amazon, Google and Meta existing.

How much do AI companies pay Wikimedia Enterprise?

No deal values or partner pricing are disclosed by Wikimedia or by any named company. The pricing page states free account limits and "Egress pricing is bespoke". New partner posts and terms changes are recorded in deal tracker updates by email.

Who owns Wikimedia Enterprise?

Wikimedia Enterprise is offered by Wikimedia, LLC, a single-member limited liability company of the Wikimedia Foundation, Inc. The Foundation is the nonprofit that operates Wikipedia.