OpenAI Says GPT-6 Astra Crosses Its Critical Cybersecurity Threshold
2026-09-03 ยท Safety & Alignment
OpenAI classified GPT-6 Astra as its first broadly deployed model to reach the Critical cybersecurity capability level under its own Preparedness Framework.
OpenAI released GPT-6 Astra on September 3, after designating it on September 1 as the company's first model to meet the Critical cybersecurity threshold in its Preparedness Framework. OpenAI says that with the right tools and access, Astra can find previously unknown security flaws and develop ways to exploit them across many well-protected systems without a person guiding each step.
OpenAI says the classification led it to strengthen protections against harmful cyber actions, whether from misuse or misalignment, and to tighten its own development and deployment of Astra: stricter isolation, checkpoint encryption, monitoring of full agent trajectories including chains of thought, and a blocking alignment evaluation before internal use. Initial access went to a limited set of organizations before a broader rollout.
"Critical" is OpenAI's own classification, not a finding by a regulator or independent body. That distinction is central for SIRDA: a developer identified a capability threshold and attached internal consequences to it, leaving open who should verify such thresholds and what happens when evaluators, governments, and developers disagree.