Exposed Hugging Face API tokens could compromise major orgs

SecurityWeek reports that more than 700 organizations deploying artificial intelligence tools, including Microsoft, Google, and VMware, could have had their accounts compromised due to the 1,681 valid Hugging Face tokens enabling large language model integration and Hugging Face repository management, which have been exposed on Hugging Face and GiHub. Numerous major firms had their repositories fully taken over by Lasso Security researchers using 77 of 655 exposed tokens that contained write permissions. Moreover, organizations with highly downloaded models could be completely accessed with the other tokens, according to the report, which also warned of training data poisoning and private model exfiltration risks as a result of the leaked tokens. "With control over an organization boasting millions of downloads, we now possess the capability to manipulate existing models, potentially turning them into malicious entities. This implies a dire threat, as the injection of corrupted models could affect millions of users who rely on these foundational models for their applications," said Lasso Security.