DataPacts

Amazon: Data Buyer Profile, Program Terms, Payouts and Reputation

Checked

At a glance

Amazon, through Amazon.com, Inc., acquires data for AI training by crawling the web with Amazonbot, whose output Amazon says may train its AI models, and by licensing content that counterparties such as Wikimedia Enterprise and Reach plc name; no Amazon program for companies that want to sell data appears in the sources checked. Two counterparty statements name Amazon or AWS as licensee, and neither discloses an amount.

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

Every figure here is a dated statement by Amazon, a counterparty or a regulator, checked 2026-11-09 unless marked, and Amazon is covered like every other buyer in this independent reference on AI training data licensing.

Each number carries its source and check date, and how DataPacts sources and dates every figure is set out in the methodology.

Amazon's Routes to Training Data: What the Record Shows

Four routes appear in Amazon's own crawler documentation, its counterparties' statements and a regulator's notice.

  • Crawling: Amazon states that Amazonbot "may be used to train Amazon AI models" (crawler page updated 2026-10-26, read 2026-11-16).
  • Paid API access: Wikimedia Enterprise named Amazon a new partner of its paid Enterprise API on 2026-01-15.
  • News content: Reach plc's half-year report of 2026-07-22 says it has "secured deals with large tech firms such as AWS".
  • Platform data: the Dutch data protection authority's notice of 2026-08-20 concerns Twitch settings for sharing data with Amazon AI.

Amazon documents page-level noarchive as its training opt-out, a tag that means "do not use the page for model training", and lists Amzn-SearchBot and Amzn-User as agents that do not crawl for generative AI training. Full rules sit with Amazonbot and the noarchive training opt-out.

Licensed content is the second route, and neither counterparty states a value, as the license table below shows. The platform route rests on one regulator record, covered under reputation below.

AWS Data Exchange is a different subject: Amazon Web Services, Inc. runs it as a marketplace where providers license data to subscribers, so a listing there is not a sale to Amazon. Its terms are in the profile of AWS Data Exchange.

Amazon Company Identity

The company on record is Amazon.com, Inc., a Delaware registrant with SEC CIK 0001018724; its cloud unit Amazon Web Services, Inc. runs the separate AWS Data Exchange marketplace. The identity table lists each attribute with its source, as of 2026-11-09.

Attribute Value Source
Legal name Amazon.com, Inc. SEC EDGAR
Registry SEC CIK 0001018724; EDGAR registrant AMAZON COM INC, Delaware SEC EDGAR
Website https://www.amazon.com amazon.com
Headquarters country United States SEC EDGAR
Founded Not recorded n/a
Type Frontier lab; Alexa named on its crawler page developer.amazon.com
Related entity Amazon Web Services, Inc., operator of AWS Data Exchange, profiled separately aws.amazon.com

No founding date is recorded because no incorporation record was checked for it. Model names other than Alexa are not described, since the sources checked hold no primary statement about them.

Amazon Program Terms and Payouts, Dated

Amazon documents no paid program for companies or contributors in the sources checked; its published terms concern how its crawlers treat a site, not payment. The table lists each term as Amazon states it or as the record shows it, as of 2026-11-09, with crawler rows re-read 2026-11-16.

Term What Amazon states or the record shows Checked Source
Company data program or application page None recorded 2026-11-09 n/a
Referral fee None recorded 2026-11-09 n/a
Payouts or amounts None disclosed by Amazon or a counterparty 2026-11-09 n/a
Crawler for training Amazonbot "may be used to train Amazon AI models" 2026-11-16 developer.amazon.com
Training opt-out Page-level noarchive: "do not use the page for model training" 2026-11-16 developer.amazon.com
robots.txt handling Allow and disallow honored; a cached copy from the last 30 days used; crawl-delay not supported 2026-11-16 developer.amazon.com
Other agents Amzn-SearchBot and Amzn-User "do not crawl content for generative AI model training"; Amzn-User "may not follow all robots.txt directives" 2026-11-16 developer.amazon.com
EU code Listed as a GPAI Code of Practice signatory; page updated 2026-10-07 2026-11-09 digital-strategy.ec.europa.eu
Art. 53 training-content summary Not located 2026-11-09 n/a
Marketplace for providers AWS Data Exchange charges providers listing fees; values are in the AWS profile 2026-11-09 docs.aws.amazon.com

Wikimedia Enterprise states no pricing and Reach plc states no value, so no per-buyer price can be derived from either statement. Other buyers publish payout ranges and referral fees, which the program terms compared set side by side.

A company looking for an Amazon application page, referral fee or license terms finds none in the sources checked. The one Amazon-run route for providers is a listing on AWS Data Exchange, where the provider sets the terms.

Amazon's Licenses on Record, by Date

The deal tracker holds two entries in which a counterparty names Amazon or AWS as licensee: Wikimedia Enterprise's partner list of 2026-01-15 and Reach plc's half-year report of 2026-07-22. The table lists them newest first, as of 2026-11-09, with values as the parties state them.

Date Counterparty What is licensed Use stated Value Source
2026-07 (disclosed 2026-07-22; signing date not disclosed) Reach plc News content; Reach: "secured deals with large tech firms such as AWS" n/s Not disclosed investegate.co.uk
2026-01-15 Wikimedia, LLC (Wikimedia Enterprise) Wikipedia and Wikimedia content via the paid Enterprise API (on-demand, snapshot, realtime); Amazon among five new partners n/s Not disclosed; no pricing enterprise.wikimedia.com

The use column reads n/s for not stated, which does not mean training is excluded. Wikimedia Enterprise states that its content spans more than 65 million articles in over 300 languages.

No statement by Amazon or The New York Times Company about an agreement between them could be read in the sources checked on 2026-11-09, so the tracker lists none. No statement by Amazon, Condé Nast or Hearst about an agreement between them was found either.

Neither entry carries a value, so no price signal exists for Amazon as a licensee, and Reach names "AWS" rather than Amazon.com, Inc., a wording kept as written. The buyer-side page on Amazon AI content licensing deals is the home for these records.

Amazon Reputation: Regulators, the EU Code and Crawler Practice

No court case over training data names Amazon in the case tracker as of 2026-11-09; the record that exists comes from a regulator's notice, the EU code's signatory list and Amazon's own crawler documentation. The table lists four records by date.

Date Body or document What it records about Amazon Status Source
2025-07-10 GPAI Code of Practice, Copyright chapter (European Commission) Voluntary code; Measure 1.3 asks signatories' crawlers to follow robots.txt per RFC 9309 Published digital-strategy.ec.europa.eu
2026-08-20 Autoriteit Persoonsgegevens (Dutch DPA) Advice to Twitch users on settings for sharing data with Amazon AI; "Training for Generative AI" on by default Advice to users, not a decision autoriteitpersoonsgegevens.nl
2026-10-07 GPAI Code signatory list Amazon listed among 22 signatories Listed digital-strategy.ec.europa.eu
2026-10-26 Amazon crawler documentation robots.txt cached up to 30 days; Amzn-User may not follow all directives Current page, re-read 2026-11-16 developer.amazon.com

Amazon states that its crawlers use a cached copy of robots.txt from the last 30 days, while RFC 9309 section 2.4 says a cached copy SHOULD NOT be used for more than 24 hours. The two statements stand side by side, and this page draws no conclusion from them.

The record does not show a Twitch ownership statement, a fine or a decision against Amazon, and the Dutch notice is advice to users. No Article 53 training-content summary for an Amazon model was located, although Article 53(1)(c) and (d) of Regulation (EU) 2024/1689 require a copyright policy and a public summary for general-purpose models.

Proceedings against other developers are tracked in the case tracker of lawsuits over training data.

Benefits, Risks, Downsides and Alternatives to Licensing Data to Amazon

Licensing to Amazon puts content with a developer that documents a page-level training opt-out and appears on the EU code's signatory list, but no Amazon terms for companies were found and no amount has been disclosed.

Benefits:

  • Amazon documents noarchive as a page-level training opt-out.
  • Amazon is on the EU code's signatory list of 2026-10-07.
  • Two counterparties name Amazon or AWS as licensee, so licensing relationships exist.

Risks:

  • Amazon states that Amazonbot output may train its models.
  • The 30-day robots.txt cache and Amzn-User's own caveat limit what a robots.txt rule alone controls.
  • The Twitch setting is on by default, per the Dutch notice of 2026-08-20.
  • No Article 53 training-content summary was located.

Downsides:

  • No Amazon program, application page or terms for companies are recorded.
  • Neither counterparty discloses a value, so no price signal exists.

Alternatives:

  • Keep the data and add noarchive to pages.
  • List data on a marketplace such as AWS Data Exchange, where the provider sets terms and pays listing fees.
  • License to other labs, or not license at all.

A holder weighing these paths finds the options compared under alternatives to selling data, including keeping the data.

Amazon Among Frontier Labs That Acquire Data

Amazon is one of the frontier labs profiled as data buyers, and it is the one whose crawler page ties a training opt-out to a page-level noarchive tag. It sits beside the other frontier AI labs as data buyers, listed here alphabetically.

More About Amazon as a Data Buyer

Amazon Connect, S3 and Comprehend in a Seller's Own Data

Amazon systems also appear on the seller side: Mercor's call-center listing names Amazon Connect among the systems an archive may come from, and its video-studio listing names Amazon S3 (Mercor's claims, checked 2026-11-09).

A holder with such an archive finds the route under selling call recordings for AI training. Amazon Comprehend detects personal data in English and Spanish (AWS documentation, 2026-11-09), one tool among those covered under PII removal and de-identification.

Amazon documents no paid contributor program in the sources checked, and Twitch's training setting is a platform default, not a payment. Paid tasks are compared under AI training jobs and platforms.

Where Each Amazon License and Crawler Is Tracked

Each record above has its own tracker or crawler page with full terms and sources.

Frequently Asked Questions About Amazon and Data Licensing

Does Amazon pay for AI training data?

No amount paid by Amazon has been disclosed by Amazon or a counterparty; Wikimedia Enterprise and Reach plc name Amazon or AWS as licensees without stating a value. The licenses on record are listed with their dates above.

Can a company license its data directly to Amazon?

No Amazon application page or terms for companies appear in the sources checked; Amazon's documented route for web content is crawling, with noarchive as the page-level training opt-out.

Does Amazon buy data through its cloud marketplace?

No. AWS Data Exchange, run by Amazon Web Services, Inc., lists providers' data for subscribers on the providers' terms, and AWS charges providers a listing fee.

Does Amazon train AI on Twitch users' data?

The Dutch data protection authority advised Twitch users on 2026-08-20 about settings for sharing data with Amazon AI, recording that the "Training for Generative AI" setting is on by default. The notice is advice to users, not a decision.