Members of the European Parliament this week adopted a comprehensive series of recommendations designed to protect copyrighted creative work from being used by artificial intelligence systems without permission or payment.

The MEPs approved the measures by 460 votes to 71, with 88 abstentions, asserting that EU copyright law must apply to all generative artificial intelligence systems available on the union’s market.

This legal framework should remain enforceable regardless of where the actual training of the artificial intelligence took place, according to the parliamentary vote.

Lawmakers insisted that the use of protected material by these systems must be fairly remunerated to safeguard the EU’s creative sector, which currently generates 6.9 per cent of the bloc’s gross domestic product.

The European Commission has been asked to examine how creators can be compensated for past use of their work, though MEPs rejected the idea of a global licence that would allow providers to train systems in exchange for a flat-rate payment.

Full transparency is a central pillar of the recommendations, as lawmakers want providers to furnish an itemised list of every copyrighted work used during the training process.

Detailed records of crawling activities for inference and retrieval-augmented generation must also be maintained, as a failure to do so could be perceived as copyright infringement and trigger legal consequences.

Should a court rule in favour of a rightsholder in such cases, the artificial intelligence providers or deployers would be required to bear all legal costs and related expenses.

The parliament has called for the creation of a new licensing market for copyrighted material, which would include voluntary collective agreements specifically tailored to individual creators and small businesses.

Rightsholders should have the clear ability to exclude their work from being used in training datasets, potentially through an opt-out list managed by the European Union Intellectual Property Office.

The news media sector requires specific protection because its work is regularly exploited by systems that divert traffic and revenues away from original outlets, the parliament stated.

Media organisations should be fully compensated for such diversions and must retain the right to refuse the use of their content for training purposes to ensure media pluralism and diversity.

The aggregation of news content must avoid the selective processing of information or any self-preferencing practices by digital gatekeepers that might benefit their own services.

Content that is fully generated by artificial intelligence should not be eligible for copyright protection, the members of parliament declared.

Digital service providers have an obligation to act against the dissemination of manipulated and generated content to protect individuals from illegal use of their likeness or data.

“We need clear rules for the use of copyright-protected content for AI training,” said Axel Voss, the Rapporteur for the Committee on Legal Affairs.

“Legal certainty would let AI developers know which content can be used and how licences can be obtained,” Voss added.

“On the other hand, rightsholders would be protected against unauthorised use of their content and receive remuneration,” he continued.

“If we want to promote and develop AI in Europe while also protecting our creators, then these provisions are absolutely indispensable,” Voss concluded.