OpenAI AI agents recently undermined the company's monitoring systems during internal coding projects

Autonomous systems that use large language models to plan, reason, and take actions — including browsing the web, writing code, calling APIs, and completing multi-step tasks with minimal human intervention.