The EU has taken a decisive step by outlawing AI systems that create sexualised deepfakes, while also postponing the rollout of its broader high‑risk AI framework.
Key Developments
- On 7 May 2026, the EU Parliament and member‑state governments voted to ban nudifier applications outright.
- The ban will be incorporated into amendments to the AI Act adopted in 2024.
- Implementation of the high‑risk AI rules has been deferred: stand‑alone AI systems now apply from December 2027 instead of August 2026, and AI embedded in products from August 2028 rather than August 2027.
- The EU executive justified the delay to protect businesses and sustain innovation, while promising continued safety oversight through other AI Act clauses.
- American AI developer Anthropic has restricted release of its powerful model Mythos, prompting EU officials to seek direct access.
- The newly empowered AI Office will begin enforcement in August 2026 and may request model access if required.
- Thirty MEPs have urged a revision of EU cybersecurity rules, citing an "emerging threat" from advanced AI tools like Mythos.
Important Facts
The ban targets AI‑generated non‑consensual sexual imagery, a response to global outrage over Elon Musk’s chatbot Grok producing such content earlier in 2026. The EU’s original timetable slated the AI Act’s high‑risk provi