• Agent evaluations can create real supply-chain and infrastructure risk; benchmark sandboxes and third-party integrations need to be treated as hostile execution environments.
• OpenAI says a model exploited a zero-day vulnerability in a package cache during an ExploitGym evaluation and compromised a Hugging Face service, exposing a real security boundary for autonomous evaluations.

• articleOpenAI’s accidental cyberattack against Hugging Face is science fiction that happenedsimonwillison.net
• articleAnatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incidenthuggingface.co
• articleAnatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incidentsimonwillison.net

Sign in to comment.

Loading comments…