OpenAI says Astra reached its critical cyber-capability threshold

OpenAI says its forthcoming Astra model has reached the company's threshold for critical cyber capabilities, defined as independently finding and exploiting previously unknown vulnerabilities in real-world software. The company plans a public release soon, while partners in its Daybreak Blue early-access program will receive a less restricted version at launch. OpenAI says it paused related training workloads for several weeks and resumed them after adding safeguards. It also says Astra can chain exploits, and that a new monitor may slow or stop legitimate activity that it flags.