Hugging Face partners with Protect AI to scan 4.47M models

AI Impact Summary

Hugging Face and Protect AI have partnered to integrate Protect AI’s Guardian scanning technology into the Hugging Face Hub, significantly expanding model security coverage. Over the past six months, Guardian has identified 4.47 million model versions and 352,000 unsafe issues, primarily through detecting vulnerabilities in model file formats like Keras and TensorFlow, as well as sophisticated obfuscation techniques. This expanded threat detection capability is crucial for developers utilizing open-source AI models on the Hugging Face Hub.