In short. AI Act Article 53 requires general-purpose AI model providers to respect copyright reservations and publish a training-content summary.

Definition: Article 53 of Regulation (EU) 2024/1689 sets obligations for providers of general-purpose AI models, models capable of a wide range of tasks and integration into different systems or applications. It requires a policy for compliance with EU copyright law and a public summary of training content, alongside documentation duties.

Key facts

The AI Act requires providers to identify and comply with copyright reservations, including through state-of-the-art technologies.
The Copyright in the Digital Single Market Directive makes its general text and data mining exception conditional on lawful access and the absence of an appropriate express rights reservation.
The Commission's training-summary guidance confirms that use of its summary template is mandatory.
The General-Purpose AI Code of Practice is voluntary and offers a route for demonstrating compliance with Article 53.

Who has these obligations, and when?

Under the AI Act's definitions and scope, a provider develops a model, or has it developed, and places it on the market under its own name or trademark. The rules cover providers placing general-purpose AI models on the Union market, including providers established outside the EU.

As of 17 October 2026, Article 113 makes the general-purpose model obligations applicable from 2 August 2025. Article 111(3) gives providers of models already placed on the market before that date until 2 August 2027 to comply.

Article 53(1)(a) requires updated technical documentation, including training, testing and evaluation, for authorities on request. Point (b) requires documentation for providers integrating the model into AI systems. The qualifying open-source exemption in Article 53(2) covers these documentation duties; the copyright-policy and public-summary duties remain applicable.

What does the copyright policy cover?

The copyright-policy provision connects the provider's compliance policy to rights reservations under the copyright directive. The exact references are:

Regulation (EU) 2024/1689
Article 53(1)(c): copyright and related-rights policy
Article 53(1)(d): public summary of training content
Directive (EU) 2019/790
Article 4(3): reservation of rights

Text and data mining means automated analysis of digital text and data to generate information, such as patterns or correlations. The directive's Article 4 provides an exception for reproductions and extractions of lawfully accessible material for this purpose. Rights holders can expressly reserve the relevant uses in an appropriate manner, such as machine-readable means for publicly available online content.

For a website owner holding the relevant rights, this reservation is the mechanism that Article 53(1)(c) requires the model provider's policy to identify and respect. Article 4(4) preserves the separate scientific-research exception in Article 3.

What can a website owner read in the training summary?

The Commission's summary guidance, updated 26 March 2026, describes information identifying the provider and model, training-content types, data sources and relevant processing. Sources include public and private datasets, scraped online content, user data and synthetic data.

The same guidance calls for publication on the provider's official website, clearly identifying the models covered, and alongside the model through public distribution channels. Where further training requires changes, it calls for updates at six-month intervals or sooner for materially significant changes.

FAQ

Does an open-source licence remove these copyright duties?

No. Under Article 53(2), the exemption concerns the documentation obligations in points (a) and (b). Copyright policy and training-content summary remain required.

Where can you look for a model's training summary?

The Commission's publication guidance points to the provider's official website and the model's public distribution channels. Check which model and version the summary covers.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer