Preparedness Framework bar, lab’s own words

Confirmed on the Astra page: OpenAI says the model meets its Critical cybersecurity threshold and can identify and develop zero-days — with stronger safeguards claimed alongside.

11 Sep 2026 · openai.com

• OpenAI writes that Astra is a significant jump in cyber capabilities and meets the Critical threshold in cybersecurity under its Preparedness Framework.
• The page says Astra’s ability to identify and develop zero-day exploits can help defenders find and patch weaknesses, and that the launching version can be used for tasks such as secure code review and patching.

Why it mattersA lab publicly labeling a broadly deployed model “Critical” for cyber is a governance signal as much as a product one.