Why AI Memory Needs Zero Trust Controls to Prevent Manipulation & Data Lock-In
What happens when AI doesn't just store what you say, but pieces together who you are? In this episode, host Shawn Rosemarin speaks with Maria Luisa Redondo Velazquez, Global CISO at TK Elevator, about the unseen security and human risks of persistent AI memory. They dive into why unlimited AI retention threatens individual identity, how "memory poisoning" turns personalization into psychological manipulation, and why security leaders must apply zero-trust principles to AI reasoning. Learn how to retain human agency over AI. Key Topics: • OT vs. IT AI Security: Why AI decisions in physical systems (like elevators) elevate risk from data loss to physical safety. • The Inference Hazard: How AI stitches harmless, isolated data points into sensitive behavioral profiles and vulnerabilities. • Zero Trust AI Memory: Why enterprises must shift from infinite memory to purpose-based, permissioned retention. • AI Memory Poisoning: How hidden prompts and commercial biases compromise model integrity and manipulate outcomes. • The Identity & Retention Trap: Why infinite digital retention threatens human evolution and creates vendor lock-in. • Human Agency by Design: Practical governance strategies for keeping humans in control of enterprise context layers. Timestamps: • 00:00 - Introduction & Security in OT vs. IT Systems • 03:24 - The Hidden Risk: Connecting Harmless Data into Sensitive Inferences • 04:56 - Context Concentration & Agent Reasoning • 06:40 - Applying Zero Trust Principles to AI Memory Retention • 08:01 - GDPR, Inferences, and Rethinking the "Right to Be Forgotten" • 11:26 - Hidden Prompts, AI Memory Poisoning & Commercial Bias • 14:17 - Hyper-Personalization vs. Psychological Manipulation • 16:07 - The Identity Trap: How Infinite Memory Prevents Human Growth • 19:14 - Enterprise Retention Discipline: Separating Memory from Authority • 22:01 - Centralization Risks & Governing the AI Context Layer • 24:04 - Memory Resilience, Poisoning Recovery & Lock-In Risks • 26:56 - Human Agency as a Core AI Design Principle