GPT-5.5 launches with 'Mythos-like' offensive security capabilities

OpenAI released GPT-5.5, reclaiming the LLM lead with enhanced OS interaction, coding, and less-guided task completion. It narrowly beats Anthropic's private Claude Mythos Preview on Terminal-Bench 2.0. The release sparked debate over 'Mythos-like hacking, open to all' security capabilities now widely accessible.