The EU Tackles AI Generating Non-Consensual Sexual Deepfakes

The European Union (EU) has taken a decisive step to ban artificial intelligences (AIs) that generate non-consensual sexual deepfakes. This decision comes in the wake of Grok's actions this winter. On X, formerly Twitter, thousands of users requested the AI to undress women and minors, and it complied without any limits. The feature sparked strong reactions worldwide, to the point that xAI ultimately decided to block it. However, the damage was already done.

Platforms and Developers Directly Targeted

The Civil Liberties Committee (LIBE) of the European Parliament has approved, in committee, a text that includes an amendment prohibiting any AI system that generates realistic images "intended to depict sexually explicit activities or the intimate parts of an identifiable physical person" without their consent. Producing and distributing sexual images without the victim's knowledge is already illegal in Europe. A 2024 directive on violence against women penalizes the use of AI to generate such visuals, while the Digital Services Act (DSA) requires platforms to remove illegal content, including sexual abuse of minors.

But this legislation goes even further: the rules would also apply to uses and users. Platforms and AI developers would be directly targeted, with the obligation to make it technically impossible, or very difficult, to generate this type of content. Today, a multitude of applications facilitate their creation.

Revision of the AI Act

The recently approved amendment is part of the AI Omnibus, a new regulation currently being adopted aimed at simplifying and adjusting the AI Act, the EU's flagship law on artificial intelligence. It notably proposes to postpone the entry into force of provisions on high-risk applications from August 2026 to December 2027, or even August 2028. This is due to complaints from industry players who fear that the legislation could stifle innovation.

It is worth noting that Elon Musk's company is not out of the woods yet. The European Union, as well as the United Kingdom, are currently investigating the generation of sexual deepfakes by Grok. It faces hefty fines.