How to Monitor Sensitive Data Exposure Across AI Systems

Summary

The video explains how shadow AI, public chatbot uploads and AI workflows can expose sensitive data, and outlines monitoring needs including data discovery, lineage, endpoint activity, policy enforcement and investigation.

Key points

• The video warns that unauthorized shadow AI projects may lack required security controls, while employees may upload sensitive information to public chatbots.
• It describes sensitive data entering AI through training data, prompts, documents, retrieval-augmented generation, policies, tools and agents.
• It recommends monitoring both AI workloads and employee use, including data transformations, file uploads and downloads, copy-paste activity and derived files.
• It emphasizes tracking data lineage from its source through AI systems to downstream destinations, even when the data changes form.
• The proposed visibility approach combines agentic-platform discovery, endpoint DLP discovery and cloud or on-premises data discovery.
• The video calls for continuous sensitive-data discovery, shared policy enforcement, risk identification and context-aware investigations.