AI Act
The world's first comprehensive AI regulation, adopted by the European Union in 2024. It governs AI systems by risk tier, imposing transparency, safety evaluation and systemic risk reporting obligations on general-purpose AI models and frontier large models. Research and development stages remain exempt, but placing a product on the EU market triggers compliance obligations.
Timeline (last 90 days)
The European Commission has sent its first AI Act information requests to OpenAI and Anthropic, marking the entry of regulation into a formal enforcement track.
The AI Act provides that if a company's response to an information request is inadequate, it can escalate to a formal infringement procedure and impose fines of up to 7% of global revenue.
The EU AI Act preserves exemptions for research, testing and development, but the Commission argues that once a frontier model reaches the EU market, the exemption no longer applies.
欧盟《AI法案》被视为AI监管领导者，但其风险分级主要针对自主行动能力较弱的旧一代模型，对能自主规划、协作和调用工具的新一代智能体覆盖不足。
China's AI regulation focuses more on immediate content-level control, while the EU AI Act places greater emphasis on ex-ante systemic risk assessment and risk classification pathways.
The EU's Digital Services Act (DSA) and AI Act impose mandatory transparency requirements on algorithmic recommendation systems.
The EU Digital Services Act (DSA) and the U.S. FTC have explicitly listed dark patterns as violations, and the EU AI Act has listed AI system transparency and explainability as mandatory requirements.
The EU AI Act classifies unlabeled deepfake content as a high-risk application subject to regulation.
All Facts (8)
The EU AI Act preserves exemptions for research, testing and development, but the Commission argues that once a frontier model reaches the EU market, the exemption no longer applies.
50%UnverifiedThe AI Act provides that if a company's response to an information request is inadequate, it can escalate to a formal infringement procedure and impose fines of up to 7% of global revenue.
50%Unverified欧盟《AI法案》被视为AI监管领导者，但其风险分级主要针对自主行动能力较弱的旧一代模型，对能自主规划、协作和调用工具的新一代智能体覆盖不足。
50%UnverifiedChina's AI regulation focuses more on immediate content-level control, while the EU AI Act places greater emphasis on ex-ante systemic risk assessment and risk classification pathways.
50%UnverifiedThe EU's Digital Services Act (DSA) and AI Act impose mandatory transparency requirements on algorithmic recommendation systems.
50%UnverifiedThe EU Digital Services Act (DSA) and the U.S. FTC have explicitly listed dark patterns as violations, and the EU AI Act has listed AI system transparency and explainability as mandatory requirements.
50%UnverifiedThe EU AI Act classifies unlabeled deepfake content as a high-risk application subject to regulation.
50%UnverifiedThe European Commission has sent its first AI Act information requests to OpenAI and Anthropic, marking the entry of regulation into a formal enforcement track.
50%