Hugging Face Breach: AI Agent Hacked Internal Systems

Hugging Face, a prominent AI platform, has disclosed a security breach. An attacker gained unauthorized access to internal datasets and service credentials. The company confirmed there's no evidence of tampering with public models, datasets, or user-facing tools. Their software supply chain was also verified as clean. The intrusion started within a data-processing pipeline. Two code-execution vulnerabilities allowed an agent to gain a foothold, access credentials, and move into internal clusters. This campaign involved thousands of automated actions across sandboxed environments. This news highlights the ongoing security challenges facing the AI industry.