EU copyright consultation weighs AI licensing options
The Commission is gathering views on AI licensing and enforcement ahead of a November 3 deadline
The European Commission opened a targeted copyright consultation on September 29, asking how European rules should respond to AI and other technological changes. Responses are due by November 3. The exercise seeks evidence for possible further measures, with copyright holders, AI providers, researchers and other affected groups invited to contribute.
Its scope extends beyond AI to piracy of live events, remuneration for music performers and producers, and copyright in scientific research. The consultation itself introduces no new legal obligations.
Licensing at different stages of AI
The questionnaire separates training uses where rights have been reserved from uses during inference and from generated output. It asks which licensing approaches fit each stage, including marketplaces for automated licensing, model contract clauses, independent mediation and collective licensing.
Enforcement options include common standards for reserving rights, a European registry for those reservations, identifying web crawlers and retaining records about content use. A separate section asks whether performers need additional protection for digital imitations of their likeness, and whether copyright is the appropriate legal route.
These are options for feedback. The questionnaire expressly says their inclusion does not mean the Commission favors them. Respondents can also support keeping current rules. They are asked to assess how possible changes would affect their businesses or activities, rather than simply pick a preferred mechanism.
Existing rules still matter
The 2019 copyright directive already distinguishes two text and data mining exceptions. Article 3 covers scientific research by research organizations and cultural heritage institutions with lawful access to the material. Article 4 provides a broader exception for reproductions and extractions of lawfully accessible material, subject to rights holders not having expressly reserved the relevant use in an appropriate way. For publicly available online content, the directive identifies machine readable reservations. Article 4 does not displace Article 3. Those conditions matter when evaluating any claim that AI training is automatically permitted or automatically infringing.
Separately, the Commission’s AI Act guidance says providers of general purpose models must implement a copyright compliance policy and publish a training content summary. Those duties apply to models placed on the EU market from August 2 2025. Providers of models placed on the market before then have until August 2 2027 to comply. These are existing requirements with their own timetable, rather than measures created by the new consultation.
A further evidence gathering stage
The Commission’s earlier call for evidence opened on May 18 and ran until June 25. It covered both a review of the 2019 directive’s practical effects and a potential targeted copyright initiative. The September consultation follows that earlier evidence gathering stage.
Participants can choose relevant sections and skip individual questions, although identification is mandatory. The form also accepts supporting documents and asks contributors to mark confidential material accordingly.
Archival photograph of the Berlaymont building by Cbliu in August 2024. Licensed under Creative Commons Attribution ShareAlike 4.0. Converted to WebP for site delivery.
