Hugging Face Servers Breached by AI Hacker; Completed Forensics in Hours Using Chinese GLM Model After U.S. API Refused Support

According to Cyberboard, Hugging Face disclosed on July 21 that its servers were recently breached by an AI hacker agent. The security team initially attempted to analyze over 17,000 attack-related logs using a U.S. commercial AI model API, but the model's safety mechanisms misidentified incident responders as attackers and refused to provide support.

Hugging Face then deployed the open-source GLM 5.2 model from China on its own infrastructure to conduct forensic analysis, which was completed successfully within hours—a task that could have taken days using traditional methods.