⚡ Quick Summary
This chapter of the Code of Practice for General-Purpose AI Models is issued under the names of Working Group 1 Co-Chair Alexander Peukert and Vice-Chair Céline Castets-Renard; no publisher or imprint is printed. Its overarching objective is to improve the functioning of the internal market and to promote the uptake of human-centric and trustworthy AI while ensuring a high level of protection of health, safety and fundamental rights, "pursuant to Article 1(1) AI Act".
The specific objectives are to serve as "a guiding document for demonstrating compliance" with Articles 53 and 55 AI Act — while recognising that adherence "does not constitute conclusive evidence of compliance" — and to enable the AI Office to assess providers who rely on the Code. Six recitals set out the legal context: the Article 53(1)(c) obligation, reservations of rights under Article 4(3) of Directive (EU) 2019/790, proportionality for SMEs, and the complementary Article 53(1)(d) training-content summary.
The operative part, Commitment 1 "Copyright policy", is developed through five measures: drawing up and implementing a copyright policy with assigned responsibilities; crawling only lawfully accessible content; identifying and complying with rights reservations, including robots.txt per IETF RFC 9309; mitigating copyright-infringing outputs; and designating a point of contact with a complaint mechanism.
🧩 What’s Covered
The chapter is organised into objectives, recitals, the legal text of Commitment 1 and five measures.
- Objectives: the overarching objective of improving the internal market, promoting human-centric and trustworthy AI and protecting health, safety and fundamental rights under the Charter, plus two specific objectives — guiding compliance with Articles 53 and 55 AI Act and enabling the AI Office to assess providers who rely on the Code.
- Recitals (a)–(f): Article 53(1)(c) AI Act requires providers placing general-purpose AI models on the Union market to hold a copyright policy and to identify and comply with reservations of rights under Article 4(3) of Directive (EU) 2019/790; Union copyright law and its enforcement are unaffected; measures should be proportionate to provider size, including SMEs and startups; the chapter complements the Article 53(1)(d) training-content summary.
- Commitment 1 — Copyright policy (legal text): restates the Article 53(1)(c) obligation and makes signatories responsible for verifying that their measures comply with Member States' implementation of Union copyright law before any copyright-relevant act.
- Measure 1.1: draw up, keep up-to-date and implement a copyright policy in a single document, assign responsibilities within the organisation, and — encouraged — publish a summary.
- Measure 1.2: when web-crawling, not to circumvent effective technological measures such as paywalls, and to exclude sites recognised by EU or EEA courts or public authorities as persistently and repeatedly infringing copyright on a commercial scale, through a dynamic EU-hosted list of hyperlinks.
- Measure 1.3: identify and comply with machine-readable rights reservations using crawlers that follow robots.txt as specified in IETF RFC 9309 and other appropriate machine-readable protocols such as asset-based or location-based metadata, with information duties towards rightsholders and an encouragement for search engines.
- Measure 1.4: implement proportionate technical safeguards against outputs that reproduce protected training content infringingly, and prohibit copyright-infringing uses in acceptable use policies, terms or model documentation, irrespective of vertical integration.
- Measure 1.5: designate a point of contact and provide a mechanism for rightsholders and their representatives, including collective management organisations, to submit substantiated complaints electronically.
💡 Why it matters?
The chapter turns an abstract AI Act duty into operational steps. Providers of general-purpose AI models placed on the Union market gain a route to demonstrate compliance with Article 53(1)(c): a documented copyright policy, crawling that respects paywalls and court-identified infringing sites, robots.txt and other machine-readable rights reservations, output safeguards, and a complaint channel for rightsholders and collective management organisations. It also states the boundary explicitly — adherence is not conclusive evidence of compliance with Union copyright law, and the Article 53(1)(d) training-content summary under an AI Office template remains a separate obligation.
❓ What’s Missing
The chapter prints no publication date, version marking, publisher or reference number, so its state in the Code's development is unclear. Several commitments are soft: signatories are only "encouraged" to publish a policy summary, to support standard-setting discussions and to avoid adverse search-engine effects. Key mechanisms depend on artefacts that do not yet exist: the EU-hosted dynamic list of infringing sites, the AI Office template for training-content summaries, and machine-readable protocols still to be developed. The text describes no assessment method, timelines, sanctions or review cycle.
👥 Best For
Legal, copyright and AI compliance teams at providers of general-purpose AI models who must implement Article 53(1)(c), and their advisers mapping crawling, rights-reservation and output-safeguard duties. It also serves policy analysts and rightsholder representatives tracking how the AI Act's copyright obligation is being operationalised in the Union.
📄 Source Details
Code of Practice for General-Purpose AI Models: Copyright Chapter. No publisher, publication date, edition or reference number is printed. The cover names Alexander Peukert (Working Group 1 Co-Chair) and Céline Castets-Renard (Working Group 1 Vice-Chair). Six pages, in English; the text extraction covered all six pages, ending with Measure 1.5. No URL for the document itself is printed in the text.
