We’re excited to announce a brand new collaboration between Hugging Face and VirusTotal, the world’s leading threat-intelligence and malware evaluation platform.
This collaboration enhances the safety of files shared across the Hugging Face Hub, helping protect the machine learning community from malicious or compromised assets.
TL;DR – Starting today, every considered one of the two.2M+ public model and datasets repositories on the Hugging Face Hub is being repeatedly scanned with VirusTotal.
Why this matters
AI models are powerful but they’re also complex digital artifacts that may include large binary files, serialized data, and dependencies that sometimes carry hidden risks.
As of today HF Hub hosts 2.2 Million Public model artifacts. As we proceed to grow into the world’s largest open platform for Machine Learning models and datasets, ensuring that shared assets remain protected is crucial.
Threats can take many forms:
- Malicious payloads disguised as model files or archives
- Files which have been compromised before upload
- Binary assets linked to known malware campaigns
- Dependencies or serialized objects that execute unsafe code when loaded
By collaborating with VirusTotal, we’re adding an additional layer of protection and visibility by enabling files shared through Hugging Face to be checked against considered one of the most important and most trusted malware intelligence databases on the planet.
How the collaboration works
Each time you visit a repository page or a file or directory page, the Hub will mechanically retrieve VirusTotal information in regards to the corresponding files. Example
Here’s what happens:
- We compare the file hash against VirusTotal’s threat-intelligence database.
- If a file hash has been previously analyzed by VirusTotal, its status (clean or malicious) is retrieved.
- No raw file contents are shared with VirusTotal maintaining user privacy and compliance with Hugging Face’s data protection principles.
- Results include metadata akin to detection counts, known-bad relationships, or associated threat-campaign intelligence where relevant.
This provides useful context to users and organizations before they download or integrate files from the Hub.
Advantages for the community
- Transparency: Users can see if files have been previously flagged or analyzed in VirusTotal’s ecosystem.
- Safety: Organizations can integrate VirusTotal checks into their CI/CD or deployment workflows to assist prevent the spread of malicious assets.
- Efficiency: Leveraging existing VirusTotal intelligence reduces the necessity for repeated or redundant scanning.
- Trust: Together, we’re making the Hugging Face Hub a safer, reliable place to collaborate on open-source AI.
Join us
In case you’d prefer to learn more about this integration or explore ways to contribute to a safer open-source AI ecosystem, reach out to security@huggingface.co.
Together, we will make AI collaboration not only open but secure by design.
