MLCommons has identified 25 distinct privacy risks associated with AI agents in a comprehensive catalog. The organization’s findings address potential vulnerabilities and concerns related to how AI agents handle, process, and protect sensitive user information.