Hugging Face, the platform that hosts AI models and datasets, disclosed last Friday that a cyberattack compromised its internal datasets and service credentials. The company said it is still investigating whether any customer or partner data was stolen during the incident.
According to a blog post from the company, an attacker uploaded a malicious dataset to the platform that exploited a security vulnerability to execute code on Hugging Face’s servers. This allowed the attacker to escalate permissions and gain broader access to internal systems. Hugging Face said it has since fixed the vulnerability and revoked and rotated the stolen credentials.
The company is urging users to rotate any access tokens or keys stored on the platform and review their accounts for suspicious activity.
Hugging Face attributed the breach to an external AI agent, which it said executed “many thousands of individual actions across a swarm of short-lived sandboxes, with self-migrating command-and-control staged on public services.” TechCrunch noted that the company did not immediately provide evidence to support this claim.
Hugging Face said its own anomaly detection systems identified the attack. When analyzing server logs, the company first turned to a frontier AI model from an unnamed commercial provider, but found the effort was blocked by that provider’s content guardrails. It ultimately used its own local large language model instead, which also avoided the need to upload sensitive attack logs to a third-party AI provider’s servers.
The company said it has reported the incident to law enforcement and engaged cybersecurity forensic specialists to investigate the breach and review its security posture.
The incident highlights the risks platforms face when attackers attempt to abuse the tools and content pipelines of a hosting service itself — rather than targeting its network perimeter directly — to gain access to sensitive internal systems.
Source: TechCrunch