EU member states approve ban on sexualized deepfake AI

What the ban is about

"Das neue Verbot soll die bisherigen Regeln ergänzen, indem es neben den Nutzenden auch die Anbieter der entsprechenden digitalen Werkzeuge ins Visier nimmt," the Council Presidency explained. The EU's AI Office is to take the lead on enforcement. "Das KI-Amt der EU soll das Verbot federführend durchsetzen," it said. The step had been classified as a formality after the agreement with the European Parliament.

The dpa specifically cited cases in which considerable damage has already been done. "Bikini-Bilder auf X oder gefälschte Nacktbilder von Stars: Sexualisierte Deepfakes haben schon viel Schaden angerichtet." Also under discussion are non-consensual images generated by the AI Grok and disseminated via the platform X. Under the current state of affairs, such non-consensual bikini images could, however, remain permitted.

Satirical caricatures are expressly excluded from the ban. "Satirische Karikaturen sollen also ausgenommen bleiben," the negotiators explained. This is intended to safeguard freedom of expression and artistic freedom without weakening protection against sexualized digital violence.

Exceptions and remaining gaps

In parallel, the Council postponed further provisions of the European AI Act. "An Teile des Gesetzes, die das Europäische KI-Amt ursprünglich bereits ab August durchsetzen sollte, müssen sich die Firmen hinter ChatGPT, Claude und Co. jetzt erst ab Dezember 2026 halten," the dpa reported. This gives providers of large AI systems a little under half a year of additional time to meet, for example, transparency and documentation obligations.

At the same time, the economy is to be relieved. "Weitere Änderungen am KI-Gesetz sollen der KI-Branche zunächst zusätzliche Pflichten ersparen und auch die Anwendung von Künstlicher Intelligenz in der Wirtschaft erleichtern," it said. Industry associations had warned in recent months about excessive bureaucracy.

Further changes to the AI Act

In addition, labeling requirements are being tightened. "Spätestens dann sollen die Anbieter KI-Inhalte etwa deutlich als solche kennzeichnen – generierte Bilder und Videos müssten sie also etwa mit Wasserzeichen markieren," the Council Presidency explained. This is intended to make it clear to consumers which content was generated by machines.

Human rights advocates and child protection organizations welcomed the ban on sexualized deepfake tools but called for improvements. They pointed out that non-consensual depictions could still remain permitted, even though they frequently serve as a gateway to digital violence. The EU Commission announced that it intends to evaluate the impact of the rule over the coming years.

Reactions and next steps

With the formal approval of the member states, the implementation phase now begins. By December 2026, national authorities must make arrangements so that penalties and withdrawal of market access can be imposed against providers of such AI applications. The EU's AI Office is also to develop guidelines for the application of the new ban.

Observers view the decision as a political signal in the fight against digital sexualized violence. The combination of a tool and provider ban, a tightened labeling requirement, and a postponement of business-related obligations is regarded as a compromise that aims to protect those affected on the one hand, while not throttling innovation on the other.