4M Models Scanned: Protect AI + Hugging Face 6 Months In

Hugging Face Blog · 1 year ago

Protect AI and Hugging Face have scanned 4.47 million model versions across their partnership since October 2024, identifying 352,000 unsafe or suspicious issues across 51,700 models. The collaboration added four new threat detection modules covering archive slip, joblib code execution, TensorFlow backdoors, and llamafile malicious code, while processing 226 million requests in the last 30 days at 7.94 milliseconds response time. Users now receive inline security alerts and vulnerability reports on the Hugging Face platform to make informed decisions about integrating models into their projects.