Static policies were built for identities that do one predictable thing. AI agents break that model, chaining across sub-agents, borrowing permissions, and deciding on the fly. In this Identity Jedi Show conversation from Identiverse, David Lee sits down with Amit Saha, co-founder and CEO of Reva, on why runtime authorization is becoming a real requirement for agentic AI, and how you secure the action instead of trying to pre-write a policy for everything an agent might do.
We get into the delegation chains that quietly escalate privilege, why identity is only about 30 percent of the visibility you need, the move from policies to guardrails plus intent, and how you make an allow, deny, or defer decision fast enough to survive at machine scale. If you are trying to let agents move fast without handing them the keys, this one is for you.
Stay in the loop The Identity Jedi newsletter breaks down non-human identity, agentic AI security, and where identity is heading, in plain practitioner language. Subscribe: [newsletter link] Subscribe to the channel for more Identity Jedi Show conversations, and drop a comment with the one thing about securing AI agents that is keeping you up at night. What would you have to change in your own stack to let an agent move fast without handing it the keys?
Podden och tillhörande omslagsbild på den här sidan tillhör
David Lee. Innehållet i podden är skapat av David Lee och inte av,
eller tillsammans med, Poddtoppen.
