Hello from DEFCON (which comes right Black Hat) - OpenAI is here with a booth promoting Daybreak, its initiative announced last month to put frontier cyber models and Codex Security in the hands of trusted defenders to find and fix vulnerabilities before attackers can exploit them.
Just a few days ago, an OpenAI briefing at Black Hat gave its first detailed account of the incident where its agents attacked OpenAI-Hugging Face infrastructure.
At the end of that Wednesday briefing, the OAI researchers emphasized that offensive AI agents will operate faster, at larger scale and with better coordination than human attackers. That means, they said, defenders will need AI systems capable of matching that speed.