Fine-Grained Authorization: The Missing Piece in Agentic AI Security
summary
Demonstrates a working RAG pipeline that uses OpenFGA + Couchbase to enforce document-level permissions before vector search, with OpenAI for embeddings and generation. A demo chat app with four user personas (CEO, lead dev, junior, salesperson) returns different answers — CEO sees financial projections, salesperson does not — proving pre-filtering blocks leakage. Claims fine-grained relationship-based policies scale to billions of checks, enable multi-tenant isolation and centralized auditing, and replace brittle role checks with graph-based permission evaluation that reduces runtime queries and latency.
AI-written summary · about
chapters (9)
- 0:00Agentic AI security risks
- 3:09OASP for LLMs
- 6:41Fine-grained authorization matters
- 10:13Authorization evolution
- 12:00OpenFGA for control
- 17:21Securing RAG with OpenFGA
- 22:57Live OpenFGA demo
- 27:13OpenFGA for AI agents
- 29:00Q&A
related talks
57mFine-Grained Authorization: The Missing Piece in Agentic AI Security
30mYour AI Agents Are Overprivileged: The Case for Fine-Grained Authorization
55mYour AI Agents Are Overprivileged: The Case for Fine-Grained Authorization
33mHow to Prevent AI Agents from Accessing Unauthorized Data
54mFine-Grained Authorization for Modern Applications
26mBeyond the Gateway: Fine-Grained Authorization for AI Agents
23mZero-Trust Architecture for Agentic AI: Securing Multi-User Access and Third-Party Integrations
40mSecuring AI RAG Pipelines with Fine Grained Authorization