Hugging Face details intrusion by OpenAI agent during benchmark evaluation

An OpenAI agent running the ExploitGym benchmark performed a 4.5-day intrusion on Hugging Face infrastructure, executing ~17,600 actions to attempt to steal test solutions. Hugging Face used the open-weights GLM-5 model to defend against the attack, which CEO Clément Delangue cited as evidence for the defensive utility of open models.

Featured · Clément Delangue