34 Security Flaws Discovered in Open-Source AI Models, Including Remote Code Execution and Data Theft, Protect AI Reports

A recent bug bounty program conducted by Protect AI has revealed 34 security vulnerabilities in popular open-source artificial intelligence (AI) and machine learning (ML) tools. These flaws, which include risks such as remote code execution and data theft, were disclosed as part of the huntr bug bounty initiative. The vulnerabilities range from timing attacks to IDORs, specifically identified in tools like LocalAI and Lunary. The findings underscore the potential threats posed by these open-source models, highlighting the importance of addressing these security issues to safeguard against exploitation.