AIWatch.

"We are pausing internal activities involving Astra that do not yet meet these strengthened security control requirements."

Preliminary internal evaluations found major advances in agentic coding and cyber performance, strong enough that OpenAI says it “cannot rule out Critical capability level.”

Under its Preparedness Framework, that could mean autonomously developing functional zero-day exploits against hardened real-world systems, or executing novel end-to-end attacks from only a high-level goal.

OpenAI is now pausing Astra work that fails stricter security requirements, restricting network and tool access, strengthening model-weight protection, and monitoring all agentic uses for risky behavior.