Wait, So Now You're Telling Me We Need FGA?
summary
Shows a concrete RAG workflow that uses OpenFGA plus the OpenAI SDK and a local vector store to prevent sensitive data leakage by filtering retrieved documents before generation. Implements a Python retriever that calls OpenFGA's batch check to ask once whether a user has viewer relation to many docs, returns only allowed documents, then generates the answer. Demo runs end to end (Docker available) and the approach scales to millions of objects.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:04Authorization for AI apps
- 3:19LLM security risks
- 5:09Authentication vs authorization
- 6:33Role-based access control
- 7:39Attribute-based access control
- 8:41Relationship-based access control
- 10:08Choosing authorization system
- 11:13Fine-grained authorization
- 13:43OpenFGA playground setup
- 14:48Securing AI apps with RAG
- 16:35Demo authorization model
- 17:40Python demo walkthrough
- 21:13Resources and OpenFGA
- 21:57Q&A
related talks
30mFine-Grained Authorization: The Missing Piece in Agentic AI Security
07mOpenFGA For Agents: Safe Delegation In 5 Minutes
37mRole-Based Access Is so Yesterday: Revolutionizing Authorization with Open FGA
57mFine-Grained Authorization: The Missing Piece in Agentic AI Security
54mFine-Grained Authorization for Modern Applications
33mFine-Grained Policies RBAC with OpenFGA
17mFine-Grained Authorization for Secure RAG With OpenSearch- 33mFine-Grained Policies RBAC with OpenFGA