All News

Industry & Advocacy News

How the EU’s New AI Act Could Impact Writers and Creators

European Union flags with a circle of gold stars on a blue field

On December 8, negotiators in the European Union reached a provisional agreement on legal regulations for the use of artificial intelligence. The law, known as the AI Act, is sweeping in scope, touching on national security, the spread of misinformation, consumer protection, privacy, law enforcement, and more. For copyright owners, the act establishes limited but potentially significant transparency obligations on companies using copyrighted works to train AI systems.

The act still needs to be formally adopted by the EU Parliament and Council, and there remain many unanswered questions about how it would apply in practice. But it could provide an important model for the United States and other countries as they consider how to implement appropriate guardrails around the growth of generative AI.

Compliance with EU Copyright Law and Transparency Obligations

While the full text of the agreement has not been released, press reports indicate that the act will require AI companies to adopt measures to ensure compliance with EU copyright law, as well as disseminate “detailed summaries” of the data used to train their models (though the criteria for those has yet to be decided). The Authors Guild applauds the inclusion of this requirement and has been working on similar legislation for robust transparency obligations in the U.S.

This critical component of AI regulation will encourage developers to work with copyright owners to license works for AI uses, instead of continuing to rely on datasets created with pirated copies of the works. While the specific details here will matter, we are pleased that the agreement takes an important first step towards ensuring transparency in AI training.

A Roadmap for U.S. Policy Makers

The act may help provide a roadmap for U.S. policymakers’ efforts to regulate AI here at home. Along with other creator groups, the Guild has been working closely with congressional offices on legislation to protect authors against unauthorized use of their works for AI training. Some AI companies have resisted such requirements by arguing that disclosing their training materials would be too burdensome given the huge volume of works ingested. Implementation of the EU law should demonstrate that transparency obligations can be crafted to appropriately balance that concern against a copyright owner’s legitimate interest in knowing whether their works are being used without permission.

Once approved, the act would not take effect until 2025 at the earliest. We will continue to monitor these developments and provide updates as more information becomes available.