The second goal of confidential AI is always to create defenses in opposition to vulnerabilities which have been inherent in the usage of ML models, such as leakage of personal information via inference queries, or https://ellasjzx674651.tblogz.com/getting-my-confidential-ai-to-work-44508901