Berlin: The Commission launched an AI copyright consultation on 29 September 2026 that asks how EU rules should treat protected works used to train artificial intelligence. Rights holders, AI developers and other stakeholders have until 3 November to respond. The exercise could change the balance between creators and technology companies, although legislation will not arrive before 2027.

The AI copyright consultation covers four topics. The first is how AI systems use copyright-protected content. The others are online piracy of live events, equitable remuneration for music performers and producers, and the application of copyright in scientific research. The questions reach beyond generative AI, but training data dominates the debate. Collective management organisations, which handle royalties for performers and producers, will watch the remuneration question closely.

The AI copyright consultation rests on two legal texts. The 2019 Copyright Directive, Directive 2019/790, contains the text and data mining exemption that many AI developers invoke when they train models on protected material. The AI Act adds transparency obligations for general-purpose AI models. The Commission asks whether these rules need strengthening through additional measures, a question it leaves open.

Parliament has already taken a position. In a resolution of 10 March 2026, MEPs proposed moving from an opt-out to an opt-in system for the use of protected works as training data, and they backed a European licensing registry. That proposal would reverse the burden, because developers would need permission before using a work instead of relying on creators to refuse.

The Commission invited a wide group to respond: rights holders, AI providers, intermediaries, collective management organisations, research institutions, consumer organisations and NGOs. Each camp in the AI copyright consultation has a different stake. Musicians and publishers want payment, developers want legal certainty for training, and researchers want to keep the room that text and data mining gives them.

After the AI copyright consultation closes, the Commission will publish a summary of the feedback and carry out an impact assessment before it drafts any legislative proposal. That sequence points to 2027 at the earliest for new rules. In the meantime, the transparency duties of the AI Act and the current mining exemption remain the operative framework.

For creators, the practical step is to respond before 3 November, since the AI copyright consultation is the first formal chance to shape the text. For developers, the coming months will show whether Brussels moves toward licensing and opt-in permissions or keeps the existing exemption. Either way, this consultation sets the agenda for the 2027 debate.