OpenAI pauses development of Astra model over cybersecurity concerns

OpenAI halted internal work on its Astra model after evaluations suggested it could identify and develop zero-day exploits in hardened systems. The model met the company's 'Critical' cybersecurity threshold, prompting the implementation of stricter security controls and universal monitoring for agentic applications.