How to Prevent AI Agents from Accessing Unauthorized Data

We're already seeing security breaches with AI Agents in the news. This is a complex problem: Imagine having N users, M Agents and O actions. How do you design permissions around that?

This talk will look at why the Google Zanzibar model of authorization which uses Relationship-Based Access Control (ReBAC) is well suited for fine-grained authorization for AI Agents at scale. The talk covers the nuts and bolts of how a Google Zanzibar system works under the hood, and how to apply it to AI Agents with techniques such as pre-filteration and post-filteration.

The talk will also include a live code demo implementing authorization for AI Agents plus RAG using Open Source tools.

Vorkenntnisse

Basic understanding of AI, LLM, RAG, and Cloud is useful.

Lernziele

You will

• learn about risks facing LLM and GenAI applications in the Enterprise
  
• get a quick primer on Authorization and an overview of Relationship Based Access Control aka ReBAC
  
• learn the basics of the Google Zanzibar whitepaper and how it works
  
• learn to know the components of a typical Agentic RAG pipeline and how to secure it
  
• get a step-by-step demonstrations of pre- and post-filtering techniques for secure data retrieval.