Skip to main content
Anthropic logo

Anthropic

EU AI Act evidence

Anthropic, PBC is an American artificial intelligence (AI) public benefit corporation headquartered in San Francisco, California

anthropic.comAI infrastructureLast verified 21 Sep 2026
Request the report Are you Anthropic?

Regulation (EU) 2024/1689

Anthropic against EU AI Act’s vendor requirements

3 provisions of EU AI Act reach the vendors you rely on. For each, the public evidence from Anthropic that may support it, with its date and source.

3 evidence found
Article 26(1) and Article 13

Use high-risk systems as the provider instructs

Deployers of high-risk AI systems must use them in accordance with the instructions for use, which Article 13 requires providers to supply with the system.

Evidence found

  • AI model and system documentation

    Evidence found

    Model documentation, system cards and training data summaries are what a deployer or integrator reads to use an AI system as intended.

    • AI model documentation"AB 2013 Training Data Summary" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5 + Fable 5 Model Documentation Form" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5 and Claude Fable 5 Training Data Summary" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5.1 & Fable 5.1 Model Documentation Form" on the trust centreas of 21 Sep 2026
    • and 11 more

  • AI governance

    Evidence found

    An AI policy, a governance framework or an ISO/IEC 42001 certificate shows the supplier manages AI risk as a system.

    • ISO/IEC 42001Vendor states ISO/IEC 42001:2023 on its trust centreas of 21 Sep 2026
    • ISO/IEC 42001ISO/IEC 42001 listed on the trust centreas of 21 Sep 2026
    • AI policy or governance framework"Anthropic Frontier Compliance Framework" on the trust centreas of 21 Sep 2026
    • AI policyListed on the trust centre as "Anthropic Frontier Compliance Framework"as of 21 Sep 2026
Article 53(1)(b)

Documentation from general-purpose model providers

Providers of general-purpose AI models must draw up, keep up to date and make available information and documentation to providers of AI systems who intend to integrate the model.

Evidence found

  • AI model and system documentation

    Evidence found

    Model documentation, system cards and training data summaries are what a deployer or integrator reads to use an AI system as intended.

    • AI model documentation"AB 2013 Training Data Summary" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5 + Fable 5 Model Documentation Form" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5 and Claude Fable 5 Training Data Summary" on the trust centreas of 21 Sep 2026
    • AI model documentation"Claude Mythos 5.1 & Fable 5.1 Model Documentation Form" on the trust centreas of 21 Sep 2026
    • and 11 more
Article 25(4)

Written terms with suppliers in the AI value chain

The provider of a high-risk AI system and a third party supplying AI systems, tools, services or components used in it must specify by written agreement the information, capabilities, technical access and assistance needed to meet the Regulation.

Evidence found

  • AI governance

    Evidence found

    An AI policy, a governance framework or an ISO/IEC 42001 certificate shows the supplier manages AI risk as a system.

    • ISO/IEC 42001Vendor states ISO/IEC 42001:2023 on its trust centreas of 21 Sep 2026
    • ISO/IEC 42001ISO/IEC 42001 listed on the trust centreas of 21 Sep 2026
    • AI policy or governance framework"Anthropic Frontier Compliance Framework" on the trust centreas of 21 Sep 2026
    • AI policyListed on the trust centre as "Anthropic Frontier Compliance Framework"as of 21 Sep 2026

  • Use of customer data for training

    Evidence found

    A published answer on whether customer data trains the supplier’s models.

    • Use of customer data to train models"Will you be using our Claude for Work conversations to train your generative models?" No. By default, we will not use your chats or coding sessions when using our Commercial products (Claude for Work, Anthropic API, etc.). If you provide feedback or otherwise explicitly opt-in to training, then we may...as of 21 Sep 2026
    • Model training on customer dataVendor answers "no" on training models with customer dataas of 21 Sep 2026

Questions buyers ask

Is Anthropic EU AI Act compliant?

CertReports does not decide that. It shows the public evidence that may support each EU AI Act requirement that reaches vendors: evidence was found or is available on request for use high-risk systems as the provider instructs (Article 26(1) and Article 13), documentation from general-purpose model providers (Article 53(1)(b)), written terms with suppliers in the ai value chain (Article 25(4)).

What should I ask Anthropic for?

Confirm the evidence below is current for the services you use, and read the contract terms each EU AI Act provision requires.

Reading for EU AI Act reviews

All articles

This maps third-party obligations to the vendor evidence that may support them. It is not legal advice and never a statement that a vendor or its customers comply; confirm scope and sufficiency with your counsel or auditor. “No public evidence” means nothing public was found at the last check. Citations link to the official text.

Requirements reviewed 21 Sep 2026. The vendor lists the evidence on its trust centre behind a request or NDA; ask for it.