Hugging Face and JFrog partner to make AI Security more transparent

Hugging Face Blog · 1 year ago

Hugging Face integrated JFrog's security scanner into its platform to detect malicious code in machine learning models alongside its existing picklescan tool. JFrog's scanner analyzes code within model weights rather than relying on pattern matching alone, catching exploits in pickle and Keras formats across hundreds of millions of files already scanned. All public models on Hugging Face Hub will now be automatically scanned for security threats when uploaded, reducing false positives and enabling safer model sharing.