Hugging Face confirms the breach of affected internal datasets and information, prompting users to take action

Hugging Face, a platform that hosts AI models and datasets, said its internal dataset and service details were compromised in a hack last week. The company disclosed the breach on Friday, but said it is still investigating whether any customer or partner data was stolen during the incident.

In a blog post, the company said the dataset uploaded to its site exploits a security vulnerability to run malicious code on its servers, allowing attackers to escalate their permissions and gain wider access to Hugging Face’s internal systems.

The company said it has withdrawn and replaced the stolen data that was accessed. It urged users to do the same with any keys stored on the platform, and review any suspicious activity on their accounts.

Hugging Face said it has fixed the vulnerability it suffered during a cyber attack. While it’s common for hackers to try to break into a company’s network using stolen employee credentials, keys, or a weak point in their security perimeter, this incident underscores the challenges companies like Hugging Face face when hackers try to abuse platforms and tools to access and steal sensitive data from within.

Hugging Face blamed the breach on an external AI agent, which performs “many thousands of individual actions within an array of short-lived, command-and-control automated sandboxes in public services.”

The company did not immediately provide proof of this claim when asked by TechCrunch.

Hugging Face said its anomaly detection detected the attack, and used an AI model to analyze the server logs that record the cyberattack.

The company said it first used a borderline AI model from a commercial provider, although it did not name the company, but found that the analysis effort was blocked by the provider’s monitors. Instead, the company used its own macro-local language model, which it says offers the added benefit of not loading sensitive attack logs on the company’s AI servers.

Security researchers have complained in the past that certain boundary models, such as Anthropic’s Mythos and Fable, are too restrictive, and prevent defenders from asking about almost anything related to cybersecurity, including defense and investigation.

Frontier AI model makers, including Anthropic, have clashed with the Trump administration over fears and concerns about the ability to use these types of cyberattacks. Anthropic was even forced to withdraw the Fable from public use after the US government imposed export controls on the model.

Hugging Face said it has notified law enforcement and engaged cybersecurity forensic experts to investigate the breach and review its security.

It is not clear whether Hugging Face did security research on its systems before it was launched. A spokesperson for Hugging Face did not respond to a request for comment Monday.