In mid-July 2026, an autonomous AI agent system breached Hugging Face's production infrastructure, exploiting vulnerabilities in the dataset processing pipeline to steal internal credentials and access internal datasets. The attack started with a poisoned dataset, exploiting a remote code execution flaw in the dataset loader and a template injection bug in dataset configuration. The agent escalated to node-level access, moved laterally across internal clusters, and harvested cloud and cluster credentials, executing more than 17,000 attacker events total. Hugging Face confirmed the credential theft but says there is no evidence “at this stage” of tampering with public models or Spaces. Hugging Face’s own LLM-based anomaly detection flagged the intrusion, and AI agents reconstructed the 17,000 actions into a timeline and credential exposure map in hours. Commercial frontier LLMs refused to analyze the exploit payloads due to safety guardrails, so the team switched to a locally hosted open-weight GLM 5.2 model to finish the investigation. Recommended actions include rotating all Hugging Face access tokens immediately, reviewing account logs for unusual activity, enabling multi-factor authentication, and contacting security@huggingface.co directly if impact is suspected.