The European Commission has opened a targeted consultation on copyright and generative AI, open until 10 November 2026

Key takeaways

• The consultation is a questionnaire, not a proposal. The Commission says the options it lists are included "for the purpose of structuring the consultation".
• On enforcement, the questionnaire asks about options including an opt-out registry, crawler identity disclosure, record-keeping, independent audit and a rebuttable presumption of use.
• On licensing, it asks about options including a licensing marketplace, model contract clauses, extended collective licensing and a remuneration right.
• A separate section asks whether specific EU rules are needed on AI-generated imitations of performers.

What is the Commission consulting on, and is it binding?

No, it is not binding. The Commission opened a questionnaire on EUSurvey. Its introduction says "The Commission is considering targeted changes to the EU's copyright framework" and that the consultation will gather input on whether future measures are needed and on possible policy options. The outcome, it says, "will feed into the preparation of a targeted initiative for a better copyright environment for European creativity and innovation". The page does not describe the form that initiative would take or when it would arrive.

The questionnaire also warns readers against treating its options as a plan. Its disclaimer says "Their inclusion should not be understood as indicating that the European Commission has taken a position in favour of a particular option". The Commission links the exercise to its call for evidence published in May 2026, and to a separate AI Strategy for the Cultural and Creative Sectors, for which it says a call for evidence was published in July 2026.

What does the generative AI section ask about training, licensing and enforcement?

The section starts from the Commission's own description of current law. It refers to the text and data mining exception in Article 4 of the 2019 DSM Directive, which applies where rights have not been expressly reserved, and to the AI Act duties on providers of general-purpose AI models to have a copyright compliance policy and publish a training data summary (Article 53(1)(c) and (d) of Regulation (EU) 2024/1689). The questionnaire says the Commission can impose fines of up to 3% of a provider's worldwide annual turnover for non-compliance. It then says rightholders struggle to enforce, license and be paid, while AI providers need legal certainty and access to high-quality data, with SMEs and start-ups facing particular difficulty.

The questions ask respondents to rank pressing issues, report their own difficulties as rightholder or AI provider, rate the impact of those barriers, and choose an EU-level approach. The four named routes are maintaining the status quo, voluntary cooperation through soft law, non-legislative measures, and new binding measures under EU copyright law. Two option lists follow, each for respondents to rank or rate.

• Enforcement mechanisms: an opt-out registry with common vocabulary and EU standards; clarifying territorial application of EU copyright rules; a duty on web crawlers and AI agents to disclose identity and purpose; record-keeping and retention by AI and dataset providers; independent audit or certification; disclosure to rightholders or collective management organisations under an NDA on a substantiated request and under certain conditions, offered as an alternative to the public disclosure of the top 10% of domain names crawled that the Commission template requires; disclosure obligations that courts or administrative authorities could impose under certain conditions in legal proceedings or mediation; and a rebuttable presumption of use, subject to specific conditions, for example where AI providers do not meet existing AI Act transparency obligations.
• Licensing and access mechanisms: a licensing marketplace; model contract clauses; a duty to negotiate in good faith; structured mediation or arbitration; extended collective licensing; fair compensation for Article 4 uses not reserved by rightholders; and a remuneration right for certain uses by generative AI providers.

The licensing table separates three uses: training of generative AI models (where rightholders have opted out from the text and data mining exception), inference-time use such as retrieval-augmented generation, and uses related to AI-generated output. Respondents can fill in at most three rows. For each option, they rate impact on their own business from strong negative to strong positive.

What does it ask about AI-generated imitations of performers?

This section is framed as a question of whether copyright is the right tool. The questionnaire states that "copyright law does not provide protection specifically against the AI generation of imitations of the performers' likeness", and points to the GDPR, the AI Act transparency rules for synthetic content and deep fakes, and national personality and image rights as other routes. It adds that some of the issues extend beyond copyright.

The questions ask how AI can support creative work in a balanced way, whether performers have had impersonations created without consent or consented to commercial exploitation but could not negotiate fair terms, and whether they have enforced rights under personality or data protection law. One asks directly whether there is a need for specific rules at EU level on performers' likeness in AI-generated content, with answers Yes, No (existing protection is sufficient) or No opinion. Follow-up questions ask what measures would help, how they would operate during a performer's life and after death, and whether copyright legislation or other legal means would be appropriate. Our article on the Danish likeness bill covers one national approach.

What else is in the questionnaire, who can respond and what are the dates?

Beyond AI, the questionnaire covers online piracy of time-sensitive content including live events, the use in the EU of sound recordings of non-EU nationals (the single equitable remuneration right after the RAAP judgment, Case C-265/19), and the research exception under the Infosoc Directive. Identification is the only mandatory section. Respondents choose which other sections to answer and may skip any question, and they can upload a supporting document and mark it confidential.

The respondent categories include individual authors, performers and their organisations, collective management organisations, AI providers, intermediary services, researchers, consumer organisations and public authorities. The language of a contribution can be English, French or German. The Commission says it may publish a summary of responses, and respondents choose between anonymous and public settings.

What should rightholders, AI developers and their counsel do now?

Decide whether to respond, because the Commission says input will shape its preparation of an initiative. Nothing in the questionnaire creates a duty for anyone who stays silent.

• Rightholders and collective management organisations: collect examples of licensing refusals, opt-out reservations and any difficulty obtaining training information, since the questionnaire asks about those barriers.
• AI providers and dataset vendors: map where copyright-protected content enters training, inference-time retrieval and output, because the licensing table asks about each stage separately. Note the cost of record-keeping or audit options before rating them.
• Performers' representatives and studios: record the contract terms now used for digital likeness and voice, and any consent and withdrawal practice.
• Counsel: treat the listed options as possibilities, not forecasts, and avoid advising a client that any of them is coming.

What we did not verify

What we opened: the Commission consultation page address (which redirects to the EUSurvey questionnaire), the questionnaire's English PDF export, and the Commission's consultations listing, all on 25 October 2026.

What we did not open: the May 2026 call for evidence, the July 2026 call for evidence on the AI Strategy for the Cultural and Creative Sectors, the European Parliament report the questionnaire mentions, the General-Purpose AI Code of Practice, the training data summary template, the privacy statement PDF (so we do not name the data controller), and the French and German language versions of the survey. The opening date comes from the consultations listing, not from the questionnaire, and the closing date comes from the questionnaire. We summarise question titles and options, not any answer.

What we refuse to claim: that the Commission has chosen any option, that it will propose legislation or on what timetable, or that any listed measure would apply to a given business.

Related briefings

• AI Regulation News hub
• Europe AI regulation tracker
• Attorneys and AI regulation
• Commission guidelines on AI Act Article 50 transparency
• Denmark copyright bill on deepfakes and likeness rights
• UK DSIT report on copyright, AI and the text and data mining opt-out
• Germany: GEMA v OpenAI, Munich ruling on AI training and copyright