Tokenization May Keep Sensitive Data Available to AI Systems

Protegrity’s tokenization approach may keep sensitive information available to AI systems while easing security concerns that limit agent deployments.

Sensitive information can spread across connected systems after entering an AI model, vector store or agent memory, making it difficult to remove. Protegrity’s approach replaces sensitive values with substitutes, allowing protected information to remain available for analytics and AI applications. Conventional encryption can restrict access to data.

Loureda said research Protegrity commissioned from Enterprise Management Associates found that 83% of surveyed leaders reported project delays. He also said 82% reported deploying AI agents with fewer capabilities after removing tools and data to address security concerns.

He urged companies to approve reusable security and governance patterns for multiple AI deployments instead of reviewing each project separately. Loureda argued that tokenization could help security teams maintain control while giving agents access to needed information.