RoboHarm benchmark finds leading AI models unreliable at rejecting dangerous robot commands
RoboHarm testing reported by The Decoder found GPT-6 Astra, Claude Fable 5.1, and MolmoAct2 unreliable at refusing dangerous robot commands.
High-Risk AI
Latest News and Analysis in High-Risk AI
High-Risk AI
EU points to OpenAI and Anthropic incidents as it argues for closer oversight of high-risk AI systems
EU officials say recent OpenAI and Anthropic incidents show why high-risk AI systems need monitoring, raising stakes for AI compliance and deployment.
European Commission Misses Critical AI Act High-Risk Systems Guidance Deadline
EU Commission fails to deliver Article 6 guidance by February 2 deadline, raising compliance uncertainty as high-risk AI rules approach August enforcement.
Ads