OpenAI pauses Astra after critical cyber risk rating

OpenAI has paused development activities for its upcoming AI model, Astra, after internal evaluations indicated it may have reached the highest cybersecurity risk level under the company's own safety framework. The company announced on August 7, 2026, that it cannot rule out Astra meeting the "Critical" threshold, marking the first time one of its models has triggered this level of concern.

The decision follows recent internal testing that showed what OpenAI described as significant advancements in Astra's autonomous coding and cybersecurity capabilities. According to the company's Preparedness Framework, a model qualifies for the Critical level if it can independently discover and develop functional zero-day exploits across all severity levels in hardened, real-world critical systems without human intervention. It can also qualify if it can devise and execute novel, end-to-end cyberattack strategies against protected targets given only a high-level objective. Previous models, including GPT-5.6 Sol, were rated at most as "High."

In response, OpenAI has implemented stricter security controls around Astra's development. The company has paused internal work on the model that does...