the 2nd objective of confidential AI is to produce defenses from vulnerabilities which might be inherent in using ML types, like leakage of private information by using inference queries, or development of adversarial https://oisimsju221288.boyblogguide.com/29804263/not-known-details-about-ai-act-schweiz