AI Training vs Copyright: The EU's New Opt-Out Debate

With copyright and AI training increasingly colliding, the European Commission has taken its next major step toward clarifying how rightsholder protections will operate under the AI Act. 

The European Commission has just closed its stakeholder consultation, which was intended to support one of the key provisions of the AI Act: the obligation for providers of AI models to identify and comply with rightsholders’ reservations of rights. Stakeholders were invited to provide their submissions before 23 January 2026. The Commission is hopeful that their industry experience will help define the state-of-the-art standards and clearer compliance obligations for general-purpose AI providers when training AI models. 

The consultation aims to establish how creators and rightsholders can indicate opt-outs that AI developers can reliably detect and respect. This would enable creators to effectively prevent their works from being used for AI model training purposes. 

GPAI providers, civil society organisations, standardisation bodies, and other interested parties were invited to share their views on the technical feasibility and practical uptake of different text and data mining (TDM) opt-out solutions. Participants could also express their interest in taking part in follow-up workshops, aimed at developing opt-out protocols that Code signatories and other providers must adhere to under the AI Act.

Now that the consultation has closed, the Commission will facilitate discussions between stakeholder groups with the aim of reaching a general agreement on a standardised list TDM opt-out protocols. This list will then be published and updated regularly, at least every two years, in line with revisions to the Code of Practice. This will ensure that it remains aligned with technological developments and evolving rights management practices. 

The closure of the consultation signals a shift in the EU’s approach to AI training practices, moving from significant theoretical debate to the implementation of practical governance solutions. These protocols aim to balance rights of creators with the operational needs of AI providers by defining how AI developers must respect rightsholders’ legally guaranteed rights of reservation. This reinforces the EU’s broader agenda of the responsible and transparent use of AI.