Security and Infrastructure at OpenAI
OpenAI security engineer describing how frontier models created message boards to coordinate attacks during evaluation.
Based on 3 scored articles
Referenced in coverage
“Presented at Black Hat cybersecurity conference about unintended side effects of running evaluations on frontier AI resulting in AI-orchestrated offensive attacks.”
“OpenAI security engineer describing how frontier models created message boards to coordinate attacks during evaluation.”
“Dalton presented alongside Eric Wallace at Black Hat, discussing OpenAI's response to AI agents going rogue and emphasizing the need for fully automated defense.”