After evaluating one of our upcoming models, Astra, we're treating it as our fir

OpenAI said on August 7 that it was treating Astra, an unreleased model, as its first "critical" model for cybersecurity, the highest level in its own risk system. The Preparedness Framework is OpenAI's internal method for tracking capabilities that could cause severe harm, with thresholds for each risk area. In the published framework (version 2, April 2025), Critical in cybersecurity means a model working with tools could find and build working zero-day exploits, attacks on flaws the vendor does not yet know about, at every severity level in many hardened real-world critical systems without human help, or plan and carry out novel end-to-end attacks on hardened targets from only a high-level goal. The framework's stated response is to halt further development until safeguards and security controls meeting a Critical standard have been specified.

The post's wording, "treating it as", signals a precautionary call rather than a settled measurement. OpenAI said it was adding controls so that Astra's further development happens safely and securely, and that it still intended to make the model broadly available and put its cyber capabilities in defenders' hands. The linked OpenAI announcement could not be retrieved for this explanation. Help Net Security reported that OpenAI described its evaluations as preliminary, said it could not rule out the Critical level, paused Astra work that did not meet stricter security requirements, and would ask government agencies and independent AI safety organizations to evaluate the model's cybersecurity capabilities before deploying it.

The precaution became a finding within a month. On September 1, OpenAI posted that Astra had reached the Critical threshold, and its safety overview for the released GPT-6 Astra calls it the first OpenAI model at that level of cybersecurity capability. The designation rests on OpenAI's own evaluations and definitions, and the August statement was a decision to act on the Critical level before those evaluations were finished.