Episode 237 - GluuFederation/identerati-office-hours GitHub Wiki
Title: Trust Before Runtime: Authority at Action Time
- Host: Mike Schwartz, Founder/CEO Gluu
- Guest: Katerina Kalimeri, Director, Cyber Security
Channels
Description
AI agent security requires three distinct controls: approve before runtime, verify at runtime, and authorize at action time. We’ll explore how agent registries, workload identity, runtime attestation and delegated authorization fit together—and why a trusted agent can still be unauthorized. Cross-industry examples will show how to tighten authority for consequential actions and record who asked, which agent acted, under whose authority and what happened.
Homework
- CSA Agentic Agent Registry Specification v1
- RFC 8693 — OAuth 2.0 Token Exchange
- OpenID AuthZEN — Authorization API 1.0
Takeaways
- ⚡
- ⚡
- ⚡
- ⚡