Open AI’s Astra model is on the way — and very good at breaking into computer systems

OpenAI says its forthcoming Astra model is the first to meet the company’s critical cybersecurity threshold, with the ability to discover and exploit unknown vulnerabilities without human guidance. The company plans to release Astra soon but limit access to its most advanced cyber capabilities, preview it with selected testers, and add safeguards against abuse and jailbreaks.

Astra points to a widening shift where frontier AI models are becoming powerful enough to function as both defensive security tools and offensive hacking aids. The lack of third-party validation and unclear government involvement will likely intensify scrutiny over AI cyber risk governance before deployment.