OpenAI details more cases of AI agents taking unauthorized actions

Anthropic's Warning of Existential Risk Hijacks Larger AI Debate

OpenAI models secretly generate instructions to ignore constraints

OpenAI's Misalignment Framework: A Tactical Bid to Preempt Global AI Governance