In short. The General-Purpose AI Code of Practice provides a voluntary route for model providers to demonstrate compliance with specified AI Act duties, including the copyright policy obligation.

Definition: The General-Purpose AI Code of Practice is a voluntary framework for AI model providers to demonstrate compliance with specified EU AI Act obligations. The European Commission's publication announcement of 10 July 2025 presents it as a tool covering transparency, copyright, and safety and security.

Key facts

The Commission's official code publication identifies Transparency and Copyright as chapters supporting Article 53 compliance, and Safety and Security as supporting Article 55 duties for models with systemic risk.
The Commission's code questions and answers explains that the code adds no obligations beyond the AI Act.
The Copyright chapter addresses the copyright policy required by Article 53(1)(c), including identifying and respecting rights reservations.

What is the code's legal role?

As of October 2026, the Commission's code publication records the Commission and AI Board's assessment that the code is an adequate voluntary compliance tool. Providers can adhere to it to demonstrate compliance with the relevant AI Act duties, or demonstrate compliance through other methods.

The Copyright chapter's objectives expressly state that adherence is not conclusive evidence of compliance with the AI Act obligations. Its recitals also preserve the application and enforcement of EU copyright law, which national courts and ultimately the Court of Justice of the European Union interpret.

The chapter identifies its legal anchor explicitly:

LEGAL TEXT: Article 53(1)(c) AI Act

That provision concerns a provider's policy for complying with EU copyright and related rights. The chapter treats the public summary of training content under Article 53(1)(d) as a complementary obligation.

What does the copyright chapter require of signatories?

The chapter's measures turn the copyright policy into operational commitments. Measure 1.1 calls for a single policy document, kept up to date, with assigned responsibility for implementation and oversight. Publishing a summary of that policy is encouraged.

Measure 1.2 covers lawful access during web crawling. Signatories commit to respect effective technological access restrictions, including those imposed by paywalls and subscriptions. They also commit to exclude websites recognised by EU or European Economic Area courts or public authorities as persistently infringing copyright commercially.

Measure 1.4 addresses outputs: it calls for proportionate technical safeguards against infringing reproduction of protected training content, alongside prohibitions on infringing use or warnings in documentation for openly licensed models. Measure 1.5 provides an electronic contact point and a mechanism for sufficiently precise, substantiated complaints from affected rights holders or their authorised representatives.

What matters for a website's rights reservations?

A rights reservation expresses a rights holder's decision to reserve a use of protected content. The Commission's copyright consultation defines text and data mining as automated analysis of digital text and data to generate information. The Copyright chapter's recital (c) describes the mining exception under Directive (EU) 2019/790 as conditional on lawful access and the absence of an appropriate express reservation.

Measure 1.3 commits signatories using crawlers for mining and model training to follow robots.txt under the Robot Exclusion Protocol, specified in RFC 9309. It also covers other appropriate machine-readable reservation protocols meeting its standardisation or adoption conditions.

For a website operator, the same measure supplies concrete checks: signatories commit to publish information about their crawlers, robots.txt features and reservation handling, and provide automatic notification of updates, for example through a web feed. Signatories that also provide or control a search engine are encouraged to prevent respect for training reservations from directly harming indexing in that engine.

FAQ

Must my business sign the code because it owns a website?

The Commission's code publication addresses voluntary adherence by general-purpose AI model providers. Owning a website alone is outside that stated signing role.

How can I raise a copyright concern with a signatory?

Look for the electronic contact point and complaints mechanism described in Measure 1.5. Your complaint should be sufficiently precise and substantiated; signatories commit to handle complaints diligently and within a reasonable time, subject to the measure's exceptions.

Sources

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