EU strikes provisional deal on landmark AI Act overhaul
What's the story
The European Union (EU) has reached a provisional agreement on its artificial intelligence (AI) Act, following nine hours of intense negotiations. The deal, which is yet to be formally approved by EU governments and the European Parliament, comes as part of a broader effort by the European Commission to simplify new digital rules. The revised AI Act entered into force in August 2024, with key provisions phased in later.
Implementation delay
Implementation of high-risk AI systems rules pushed back
The provisional agreement also pushes back the implementation of rules for high-risk AI systems, such as those involving biometrics or related to critical infrastructure and law enforcement. The new deadline is now December 2, 2027, instead of the earlier date of August 2 this year. This change comes in response to complaints from businesses about overlapping regulations and red tape that hinder their competitiveness against US and Asian rivals.
Exclusion clause
Machinery excluded from AI Act
The provisional agreement also excludes machinery from the AI Act, as it is already covered by sectoral rules. This decision was taken after pressure from businesses. The deal also includes a ban on AI practices that create unauthorized sexually explicit images, a move in response to such content generated by Elon Musk's xAI chatbot Grok on X and sexually intimate deepfakes produced by Grok.
Watermark requirement
Mandate for watermarking AI-generated output
The provisional agreement also mandates watermarking of AI-generated output, effective from December 2. This comes as part of the EU's broader effort to address concerns about the impact of technology on children, workers, companies, and cybersecurity. Despite these changes, the AI rules are still considered some of the strictest in the world.