We are pleased to announce a new collaboration between Hugging Face and VirusTotal, the world’s leading threat intelligence and malware analysis platform. This collaboration strengthens the security of files shared across Hugging Face Hub and helps protect the machine learning community from malicious or compromised assets.
TL;DR – Starting today, all of the 2.2 million+ public model and dataset repositories on Hugging Face Hub will be continuously scanned by VirusTotal.
why is this important
AI models are powerful, but they are also complex digital artifacts that include large binary files, serialized data, and dependencies with hidden risks. Currently, HF Hub hosts 2.2 million public model artifacts. As we continue to grow into the world’s largest open platform for machine learning models and datasets, ensuring the safety of our shared assets is essential.
Threats come in many forms.
Malicious payload disguised as a model file or archive Files that are compromised before being uploaded Binary assets linked to known malware campaigns Dependencies or serialized objects that execute insecure code when loaded
By working with VirusTotal, we are adding an extra layer of protection and visibility by allowing files shared through Hugging Face to be checked against one of the world’s largest and most trusted malware intelligence databases.
How collaboration works
When you access a repository page, file, or directory page, the hub automatically retrieves VirusTotal information about the corresponding file. example
Here’s what happens:
Compare file hashes to VirusTotal’s threat intelligence database. If a file hash has been previously analyzed by VirusTotal, its status (clean or malicious) is retrieved. To maintain user privacy and compliance with Hugging Face’s data protection principles, raw file contents are not shared with VirusTotal. Results include metadata such as number of detections, known bad relationships, or associated threat campaign intelligence.
This provides valuable context before users or organizations download or integrate files from the hub.
Benefits for the community
Transparency: Users can see if their 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 prevent the spread of malicious assets. Efficiency: Leveraging existing VirusTotal intelligence reduces the need for repetitive or redundant scans. Trust: Together, we’re making Hugging Face Hub a safer, more trusted place to collaborate on open source AI.
would you like to join
If you would like to learn more about this integration or discuss how you can contribute to a more secure open source AI ecosystem, please contact us at security@huggingface.co.
Together, we can design AI collaboration to be not only open but also secure.
