Proactive AI privacy – Leveraging PETs throughout the AI life cycle

Privacy challenges exist at each stage in the artificial intelligence life cycle. Collecting rich, quality training data at scale must be consistent with principles of user consent and data minimization; model memorization raises privacy risk at training time; membership inference attacks can create privacy problems during deployment. Agentic AI adds complexity as agents autonomously access systems and use tools across multiple sessions. Learn how privacy-enhancing technologies like differential privacy, federated learning and synthetic data generation are implemented at each stage in the AI life cycle to proactively address privacy risks and meet data processing goals. Practitioners on the front lines of PETs integration share strategies for effective deployment and performance evaluation. The session will also explore cutting-edge research and standardization efforts to address PETs limitations and lower barriers to use.

• View presentation