Unfortunate news emerged late last Friday as AI startup Hugging Face revealed that its security team had detected unauthorized access to its platform, Spaces. Spaces is a platform used to create, share, and host AI models and resources. The intrusion specifically targeted Spaces secrets, which are private pieces of information that act as keys to unlock protected resources such as accounts and tools. Hugging Face expressed suspicions that a third party may have gained access to some of these secrets without authorization.
As a precautionary measure, Hugging Face has revoked a number of tokens associated with the compromised secrets. Tokens are used to verify identities, and users whose tokens have been revoked have already received email notifications. The company is advising all users to refresh any key or token and consider switching to fine-grained access tokens, which are believed to be more secure.
The exact impact of the potential breach on users and apps remains unclear. Hugging Face is actively working with external cybersecurity forensic specialists to investigate the incident and review its security policies and procedures. The company has also reported the incident to law enforcement agencies and data protection authorities. In their blog post, Hugging Face expressed regret for any disruption caused by the incident and emphasized their commitment to using this opportunity to reinforce the security of their entire infrastructure.
A spokesperson from Hugging Face acknowledged that cyberattacks have been on the rise in recent months, possibly due to the company’s growing usage and the increasing mainstream adoption of AI. However, determining the precise extent of the compromised spaces secrets is technically challenging.
This potential hack comes at a time when Hugging Face is already facing scrutiny regarding its security practices. In April, a vulnerability was discovered by researchers at cloud security firm Wiz, allowing attackers to execute arbitrary code during the build time of a Hugging Face-hosted app, enabling them to examine network connections from their own machines. Additionally, security firm JFrog found evidence earlier in the year that code uploaded to Hugging Face had covertly installed backdoors and malware on end-user machines. Another security startup, HiddenLayer, identified ways in which Hugging Face’s supposedly safer serialization format, Safetensors, could be exploited to create compromised AI models.
In an effort to enhance its security measures, Hugging Face recently announced a partnership with Wiz. The collaboration aims to utilize Wiz’s vulnerability scanning and cloud environment configuration tools to improve security not only on the Hugging Face platform but also across the wider AI and ML ecosystem.
While Hugging Face faces challenges with regard to its security practices, it is commendable that the company is taking proactive steps to address these issues. By partnering with cybersecurity experts and actively working to enhance its security infrastructure, Hugging Face is demonstrating its commitment to protecting the privacy and data of its users. As the company continues to grow and AI becomes more mainstream, it is crucial for platforms like Hugging Face to prioritize security and remain vigilant against cyber threats.
