Responding to the next frontier of critical cyber capabilities
OpenAI releases preliminary cybersecurity evaluations for its Astra system
OpenAI is publishing preliminary cybersecurity evaluations for Astra alongside steps to strengthen safeguards and security controls. This is a notable transparency move signaling OpenAI's recognition that its frontier models carry meaningful cyber risk. It fits a broader pattern of labs proactively disclosing safety posture before regulatory pressure forces it.