03:34 OpenAI GPT-5.5 matches Claude in cyberattack potential, AI Security Institute reports Ryan Torres
02:23 UK AI Security Institute finds Anthropic's Mythos model excels at multi-step cyberattacks Ryan Torres
