EU negotiators reach provisional deal to ban AI tools used to create sexualized deepfakes
The facts
- EU member-state representatives and European Parliament negotiators reached a provisional agreement on changes to the EU AI Act on May 7.
- The agreement includes a ban on AI applications or systems used to create sexualized deepfakes.
- The ban is intended to cover content involving child sexual abuse material as well as non-consensual sexualized content involving adults.
- Before the changes can take effect, the agreement still must be formally approved by the European Parliament and the Council of member-state governments.
- The new prohibition is part of a wider package that also simplifies or delays other AI rules in the EU.
- If the reform is adopted, enforcement of the ban would be led by the EU AI Office.
- Multiple reports say the ban would apply from December 2, 2026, giving companies time to bring systems into compliance.
Context
What content is the EU trying to stop with this ban?
The agreement targets AI-generated sexualized deepfakes, including content depicting child sexual abuse and explicit material involving adults created without their consent Independent,Schwarzwälder Bote,Hindu.
Is the ban already law?
No. Negotiators reached a provisional deal, but it still needs formal approval from the European Parliament and the Council before it can enter into force Reuters,tagesschau.de,newsORF.at.
What happens next if the deal is approved?
Reports say the ban would be enforced by the EU AI Office and would apply from December 2, 2026, while companies would need to make their systems compliant by that date Südtirol News,onvista.de,Independent.
Common ground, and where they split
- Common ground
- Sexualized deepfakes are being treated as a distinct, clearly harmful use of AI that warrants an explicit ban, covering both child sexual abuse material and non-consensual sexualized content involving adults rather than being left to general AI rules.
- Where Left and Right split
- Less a disagreement than a question of emphasis: protecting people from a rights and public-safety abuse, versus setting a targeted legal boundary around one harmful category while avoiding a blanket approach to AI.
How left and right read it
The core issue here is protection against a form of abuse that can target both children and adults without consent. A ban on AI systems used to create sexualized deepfakes treats this as a public-safety and rights problem, not just a technical dispute, even as it sits inside a broader package that also delays or simplifies other AI rules.
What matters most is drawing a firm boundary around a clearly harmful use of AI while still moving through the EU’s formal approval process before the change takes effect. Folding the ban into a wider reform package also underscores that lawmakers are trying to set limits on one category of conduct without treating every AI application the same way.
The receipts — 100 sources
Wire services (4)
Other coverage (96)
A news briefing built to show both sides.
The day’s biggest stories in one short brief, built to show both sides and the competing values behind the headlines. Free in your inbox.