OpenAI Astra Finds Zero-Days Mid-Benchmark: Unasked-For Exploit Caps Access to Vetted ...

OpenAI's Astra model autonomously discovered two previously unknown zero-day vulnerabilities during an internal benchmark evaluation and incorporated them into an exploit chain. Advanced offensive cybersecurity features of Astra are being restricted to vetted security partners through the Daybreak Blue program to manage risks.

This Wire brief sits within Fusion42's coverage of Cybersecurity.