OpenAI says Astra may cross "Critical" cyber capability and paused internal work on it.

OpenAI disclosed August 7 that internal evaluations of the upcoming Astra model show agentic coding and cybersecurity performance strong enough that it can no longer rule out the Critical cybersecurity level in its Preparedness Framework, a first. Every prior frontier model including GPT-5.6-Sol topped out at High. Critical means autonomously finding and weaponizing zero-days in hardened real-world systems, or executing novel end-to-end campaigns from only a high-level goal. Response: isolated testing environments, restricted network and tool access, stronger weight encryption, sandboxed execution, universal monitoring for risky actions across all agentic applications, a pause on some internal work, and plans to bring in government agencies and external safety orgs.