Protect AI and Hugging Face have reached a significant milestone in their collaborative effort to secure the open-source artificial intelligence ecosystem. Six months after launching their joint security initiative, the organizations have successfully scanned over 4 million machine learning models hosted on the Hugging Face platform.
Identifying Structural Vulnerabilities
The scanning process utilizes Protect AI's specialized security tools to detect 'pickles' and other potentially malicious file formats that can lead to remote code execution (RCE) vulnerabilities. By analyzing the structural integrity of these models, the partnership aims to prevent supply chain attacks within the AI development lifecycle.
Strengthening the AI Supply Chain
As the adoption of open-source AI continues to accelerate, the security of pre-trained models has become a primary concern for enterprises. This initiative provides developers with greater transparency regarding the safety of the assets they integrate into their applications. The ongoing monitoring helps establish a baseline for security standards in an industry that is rapidly evolving and increasingly targeted by sophisticated cyber threats.
