Srinivas Sandiri

Talk Title:

From Identity to Assurance: Securing the Enterprise AI Agent Execution Path

Abstract:

Enterprise AI is moving from generating answers to taking actions across data, tools, APIs, workflows, and downstream systems. This shift creates a security challenge that identity alone cannot solve: what effective authority can an AI agent exercise across the full execution path, and how can organizations demonstrate that this authority remains appropriately bounded?

This keynote presents a vendor-neutral architecture approach for tracing identity and effective authority from human or system initiators through AI agents, tools, APIs, and enterprise applications. It connects least privilege, deterministic action controls, human oversight, observability, testing, and evidence into an assurance model for increasingly autonomous systems. The goal is to move beyond prompt-level guardrails toward AI systems whose actions are authorized, constrained, observable, and verifiable.

Profile:

Srinivas Sandiri is a Principal Security Architect with over two decades of experience in enterprise architecture, cybersecurity, and large-scale technology transformation. His work focuses on enterprise security architecture, AI security and governance, identity and access management, secure integrations, cloud and SaaS platforms, and the security architecture of AI-enabled enterprise systems. He has contributed to industry publications, technical conferences, professional judging activities, and architecture initiatives involving complex enterprise environments. His current areas of interest include secure enterprise AI adoption, agentic AI security, effective authority, deterministic controls, and evidence-based security assurance.