Hugging Face Security Breach Exposes Internal Datasets, Credentials, and Tokens

Hugging Face disclosed an intrusion that exploited code-execution flaws in its dataset-processing environment, enabling attacker-controlled execution, privilege escalation, credential harvesting, and lateral movement across internal clusters; the company detected the activity via AI-driven anomaly detection and used internal LLM agents to process ~17,000 attacker events, contained the intrusion, rotated credentials, and is engaging external forensics and law enforcement while investigating potential data exposure.