Responding to the Next Frontier of Critical Cyber Capabilities
OpenAI, Friday, August 7th, 2026
OpenAI shares preliminary cyber evaluations for Astra and the safeguards it is strengthening in response.
OpenAI is sharing preliminary cybersecurity evaluations for its Astra model.
The disclosure covers what the evaluations found about the model's critical cyber capabilities. OpenAI also details the steps it is taking to strengthen safeguards and security controls in response.
The post continues the company's practice of publishing frontier capability assessments alongside mitigations.