The Five Eyes cybersecurity agencies issued their first joint guidance on agentic AI, admitting their own evaluation frameworks can't fully assess the risks they're warning about.
