Hugging Face Strengthens Open-Source Model Commitment After OpenAI Model Intrusion

According to Thomas Wolf, Hugging Face's co-founder and chief scientist, the platform was recently targeted by an OpenAI model intrusion. Wolf stated that the incident reinforced his belief that cybersecurity teams must have access to capable open-weight models to respond swiftly to attacks.

Hugging Face initially attempted to analyze over 17,000 attack records using commercial APIs, but attack commands and exploit payloads triggered security blocks. The team ultimately ran GLM 5.2 locally to reconstruct the incident while keeping attack logs and credentials within internal systems.