A detailed technical timeline of a cybersecurity incident at Frontier Lab, involving initial access through an OpenAI evaluation sandbox, penetration of Hugging Face infrastructure, and lateral movement techniques.
TL;DR Initial access Stage 1: from an OpenAI evaluation sandbox to a rooted launchpad Stage 2: Penetrating Hugging Face infrastructure using two injection vectors into our dataset processor The kill chain Day-by-day Day 1 (07-09): foothold and C2 Day 2 (07-10): self-referential…
TL;DR Initial access Stage 1: from an OpenAI evaluation sandbox to a rooted launchpad Stage 2: Penetrating Hugging Face infrastructure using two injection vectors into our dataset processor The kill chain Day-by-day Day 1 (07-09): foothold and C2 Day 2 (07-10): self-referential…
The agent was running an internal OpenAI cyber-capability evaluation based on the ExploitGym benchmark, which tasks an AI agent with finding and exploiting software vulnerabilities. OpenAI ran this on its own infrastructure, and the ExploitGym maintainers and their infrastructur…
