EU AI Act Officially Takes Effect With Staggered Compliance Timeline
The European Union's landmark AI Act is now in force, kicking off a phased rollout of regulations based on application risk levels. The first rules banning specific high-risk AI uses will take effect in six months.
The European Union’s AI Act officially takes effect today, initiating a phased approach to regulating artificial intelligence across the bloc. The law categorizes AI applications into risk tiers, leaving most low-risk systems unregulated while imposing strict obligations on high-risk uses like biometrics and employment software. The first major deadline arrives in six months, when authorities enforce bans on prohibited AI practices such as law enforcement's use of remote biometrics in public spaces.
Developers of high-risk AI systems face rigorous requirements, including pre-market conformity assessments and mandatory registration in an EU database if used by the public sector. Limited-risk tools like chatbots and deepfake generators must meet transparency standards to prevent user deception. The legislation also establishes steep financial penalties that scale with the violation, reaching up to seven percent of global annual turnover for deploying banned AI applications.
The regulation introduces specific rules for general-purpose AI (GPAI) developers, applying lighter transparency requirements to most models but demanding stricter oversight for the most powerful systems. Models trained using over 10^25 FLOPs of computing power face additional risk assessment and mitigation mandates. While member states handle enforcement for general AI rules, the EU directly oversees compliance for these powerful GPAI models as the full regulation comes into effect by mid-2026.