OpenAI's GPT-5.5 Can Improve AI Itself

OpenAI has launched GPT-5.5, described as a new class of intelligence capable of handling complex research tasks with minimal guidance, including improving AI systems themselves. Co-founder Greg Brockman called it a foundation for future computer work, highlighting its strength in agentic coding and autonomous task completion. The model includes OpenAI's strongest safety measures yet, targeting misuse in bio and cyber domains. The release comes amid intensifying competition, with rival Anthropic restricting its Claude Mythos model over hacking concerns. OpenAI says GPT-5.5 marks one of the clearest steps yet toward AI that can accelerate its own research and development.