Hugging Face has disclosed a security incident involving Spaces, its platform for creating, sharing and hosting AI models and resources. The company said its security team detected unauthorized access earlier this week and is investigating whether private secrets connected to Spaces were exposed.
The company has not said how many users or apps may have been affected. Its response so far centers on token revocation, user notifications and a broader security review.
What Hugging Face says happened
According to Hugging Face, the incident involved Spaces secrets. These secrets are private pieces of information that function like keys, giving access to protected resources such as accounts, tools and development environments.
Hugging Face said it has suspicions that some of those secrets could have been accessed by a third party without authorization. That wording matters: the company is describing possible exposure, not a fully quantified breach with a confirmed impact count.
As a precaution, Hugging Face revoked a number of tokens contained in those secrets. Tokens are used to verify identity, so revoking them is a way to reduce the risk that an exposed credential can continue to be used.
Users whose tokens were revoked have already been notified by email, according to the company. Hugging Face is also recommending that all users refresh any key or token and consider moving to fine-grained access tokens, which the company says are more secure.
Why Spaces secrets matter
Spaces is part of the infrastructure that lets people build and share AI-powered work. In that context, secrets are not ordinary settings. They can connect an app or project to protected accounts, tools and development environments.
That is why even the suspicion of unauthorized access can trigger a broad reset. If a secret is exposed, the safest assumption is that the key may no longer be private. Refreshing keys and tokens breaks that possible chain of access.
The company has not provided a number for affected users or apps. A spokesperson told TechCrunch that it is technically difficult to know how many Spaces secrets may have been compromised.
For users, the practical takeaway is straightforward:
- Check whether Hugging Face sent an email notice about revoked tokens.
- Refresh keys or tokens connected to Spaces.
- Consider switching to fine-grained access tokens.
- Review connected accounts, tools and development environments that depend on those secrets.
The investigation is still moving
Hugging Face said it is working with outside cyber security forensic specialists to investigate the issue. The company also said it is reviewing security policies and procedures.
In its post, Hugging Face said it had reported the incident to law enforcement agencies and Data protection authorities. The company also apologized for the disruption and said it would use the incident to strengthen the security of its infrastructure.
In an emailed statement to TechCrunch, a Hugging Face spokesperson linked the pressure to the rising profile of AI. The spokesperson said cyberattacks have increased significantly in recent months, probably because Hugging Face usage has grown significantly and AI is becoming more mainstream.
That explanation does not identify who was behind the incident. It does, however, frame the event as part of a broader increase in attention on AI platforms that host models, datasets and apps.
A wider security spotlight on Hugging Face
The Spaces incident comes while Hugging Face is already under closer examination for security practices. TechCrunch describes Hugging Face as one of the largest platforms for collaborative AI and data science projects, with over one million models, data sets and AI-powered apps.
In April, researchers at cloud security firm Wiz found a vulnerability that has since been fixed. The flaw could have allowed attackers to execute arbitrary code during the build time of a Hugging Face-hosted app and examine network connections from their machines.
Earlier in the year, security firm JFrog uncovered evidence that code uploaded to Hugging Face covertly installed backdoors and other types of malware on end-user machines. Security startup HiddenLayer also identified ways that Safetensors, Hugging Face's serialization format described as ostensibly safer, could be abused to create sabotaged AI models.
Hugging Face recently said it would partner with Wiz to use vulnerability scanning and cloud environment configuration tools. The stated goal was to improve security across the platform and the AI/ML ecosystem at large.
What the incident signals
The clearest immediate signal is that credential hygiene is central to AI platform security. Models and apps may get the attention, but secrets, tokens and access controls determine what connected systems can be reached.
Hugging Face has not disclosed a confirmed count of compromised secrets, users or apps. That uncertainty is part of the risk. When a platform cannot immediately determine the scope, revoking tokens and asking users to refresh credentials becomes a defensive baseline.
For Hugging Face, the incident adds urgency to its already public security work. For users, the next step is less abstract: treat old Spaces-related keys and tokens as credentials that may need to be replaced.