Hugging Face Discloses Breach Carried Out End-to-End by an Autonomous AI Agent Swarm

Hugging Face said an autonomous AI agent system breached part of its production infrastructure, in what the company describes as an attack executed end-to-end by machine agents rather than a human operator at the keyboard. In a security disclosure published on its blog, the AI platform said the incident began when a malicious dataset exploited two code-execution weaknesses in its data-processing pipeline, then escalated into privilege elevation, credential theft and lateral movement across internal clusters over a weekend.