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

OpenAI is preparing to release its new "Astra" model, which the company describes as the first large language model to meet its "critical cybersecurity threshold." OpenAI stated that while Astra will be available soon, access to its most advanced cybersecurity capabilities will be limited. Internal testing revealed that Astra is capable of identifying and exploiting unknown security flaws in computer systems without human guidance. This capability mirrors concerns previously raised by Anthropic regarding its "Mythos" model, prompting OpenAI to implement similar safety precautions. Notably, Astra achieved a perfect score on "ExploitBench," an evaluation measuring an LLM's ability to hack into known system vulnerabilities. In a modified version of this test, the model successfully discovered and exploited two zero-day vulnerabilities. To mitigate risks, OpenAI has begun improving the model's ability to detect abuses and prevent jailbreaks. Furthermore, the company is identifying high-risk accounts to restrict their access and will deploy the model with additional chain-of-thought monitoring to detect and prevent harmful behavior.