European Union lawmakers and countries agreed on Thursday to ban artificial intelligence systems generating sexualised deepfakes, after global outrage over non-consensual nudes produced by Elon Musk's chatbot Grok. The new ban will be included in changes to the EU's comprehensive rules on AI, adopted in 2024.

The EU negotiators also agreed to delay the implementation of high-risk AI rules, concerning models deemed potentially dangerous to safety, health or citizens' fundamental rights, from August 2026 to December 2027 and from 2027 to August 2028 for stand-alone and embedded AI systems, respectively.

EU centrist lawmaker Michael McNamara said the EU has drawn a red line, stating that AI must never be used to humiliate, exploit or endanger people.

The EU executive proposed the amendments last year to help businesses and avoid stymying innovation, and still hopes to steer the technology's safe development via other provisions of the AI Act.

The EU's AI regulation office will have unique access to providers' internal safety and security practices, and will ensure to receive model access if needed, starting from August 2026. The European Parliament will now finalize the agreed text, with the aim of adopting the revised AI rules in the coming months.