Unpatched Critical Vulnerabilities in AI Models Pose Risks of Unauthorized Access, Information Theft, and Model Poisoning

Researchers at Protect AI and independent security experts have identified critical vulnerabilities in the infrastructure used by AI models on the Huntr Bug Bounty Platform, including high-severity bugs in platforms such as Ray, MLflow, ModelDB, and H20 version 3, potentially leading to server takeover and unauthorized access to AI models, with concerns of intellectual property theft and financial gain for cybercriminals.