Setting Security Boundaries for Agentic AI | PlainID's Gal Helemski With скачать в хорошем качестве

Setting Security Boundaries for Agentic AI | PlainID's Gal Helemski With

In this session, Gal Helemski joined ‪@kuppingercole‬'s John Tolbert to discuss why traditional access models fall short for agentic AI, and why authorization must become a continuous, real-time security decision. You’ll hear how organizations can set practical security boundaries across the full AI flow, including: Prompt guardrails to block unauthorized requests, Data retrieval guardrails to control what information enters the flow, MCP and tool guardrails to govern what services agents can invoke, Output guardrails to mask and filter sensitive responses, Why inherited human access and static roles are not enough, How dynamic, context-aware authorization helps enforce Zero Standing Privileges. We trust this conversation will be especially relevant for CISOs, security architects, IAM leaders, and teams planning or piloting agentic AI in the enterprise. PlainID’s approach centers on controlling what every identity, human and non-human, can access, do, and expose across the full AI flow. Learn more here: https://www.plainid.com/ai/