The European Union and European Parliament have reached a provisional agreement to amend the bloc's landmark AI Act, including a ban on using artificial intelligence to create unauthorized sexually explicit deepfakes and a delay in implementing certain high-risk AI rules.

Under the deal, obligations for high-risk AI systems—covering biometrics, critical infrastructure, education, employment, law enforcement, and border management—will be postponed to December 2, 2027, from the original deadline of August 2025.

Machinery and equipment have been excluded from the AI Act, as they are already subject to sectoral safety rules. Critics view this as a concession to business pressure, watering down the original regulatory ambition.

Mandatory watermarking of AI-generated content will take effect on December 2, 2025. Marilena Raouna, Cyprus's Deputy Minister for European Affairs, claimed the agreement reduces administrative costs and ensures legal certainty, though skepticism remains about enforcement.

The most notable provision is a ban on creating unauthorized sexually explicit deepfakes, including images, video, and audio. Companies have until December 2, 2025, to comply. The ban also explicitly covers child sexual abuse material.

The EU's move allegedly responds to incidents where users used Elon Musk's Grok chatbot to generate and spread millions of AI-generated sexual images of women and children. The agreement now requires formal approval from the European Parliament and EU governments, which is considered a formality.

Source: www.dw.com