Building Trustworthy AI Authentication & Authorization for AI Agents
summary
Shows concrete engineering choices: use an Alzero token vault so servers fetch short lived tokens just in time instead of exposing refresh tokens to agent memory, use SBA back channel and push via Guardian for human approval on risky actions, and enforce access with open FGA fine grained authorization so per document relations block unauthorized transfers. Demo is a banking agent that approves a user transfer but denies an attempted corporate card theft in real time.
AI-written summary · about
chapters (9)
- 0:00Trustworthy AI security intro
- 2:29Security risks in agentic AI
- 5:18Authentication challenges
- 8:48Token management with vault
- 10:59Human-in-the-loop for sensitive actions
- 12:24Client-initiated back-channel auth
- 14:52Fine-grained authorization with OpenFGA
- 19:31Demo: secure banking agent
- 23:27Implementation resources
related talks
29mHow Auth0 Secures AI Authentication & Authorization for AI Agents
17mDon't give your agent the master key
40mIn AI We Trust? Securing the Future, One Agent at a Time
55mYour AI Agents Are Overprivileged: The Case for Fine-Grained Authorization
19mSecuring Agents with Open Standards
29mBuilding Trust in AI Agents
30mFine-Grained Authorization: The Missing Piece in Agentic AI Security
01h 22mIdentity for AI Agents