Architecting Trusted Agents: Turning Knowledge Graphs into Secure Policy Engines

You gave your AI agents access to the knowledge graph. Now they can answer anything, grounded in real data, no hallucinations. Problem solved, right? Not quite. What happens when someone asks the agent to summarize confidential HR files? Or pull the CEO's salary? What if an agent deletes a critical asset and you have no idea why? And how does any of this comply with GDPR or CCPA? Security is the part most teams skip when building knowledge-based AI. And when they do think about it, they reach for role-based access control, which quietly turns into a mess of role explosion and over-permissioning. That's a bad combination with AI, where behavior is dynamic and hard to predict. In this session, you'll learn how to turn your own graph into a policy engine that your agents can actually trust. We cover: Why connected data raises the security stakes for agentic AI Why RBAC breaks down and what to use instead How to apply your graph as security context Real standards you can use today: MCP, A2A, OAuth2, and the new AuthZEN authorization standard

Guest: Alex Babeanu ( https://www.linkedin.com/in/ababeanu/ )

#neo4j #graphdatabase #agenticai #knowledgelayer #knowledgegraph #graphrag #AIAgents