OpenAI’s Astra Becomes First AI Model with ‘Critical’ Hacking Abilities

The jump to autonomous exploit chaining marks a qualitative shift in AI risk, moving from assistance to operation. This matters because 'critical' designation implies capabilities that could outpace existing defensive automation, potentially redefining the threat landscape for enterprises and security teams relying on patch cycles. With access gated to testers, the immediate watch is whether discovery rates of zero-days in the wild correlate with this tool's deployment window, and how OpenAI's disclosure policies handle any findings.