How to Discover, Connect & Govern AI Agents | Okta CEO Todd McKinnon & SVP AI Security Harish Peri
📅 Jun 18, 2026Okta29 MIN302 VIEWS26 SEGMENTS · 3 SPEAKERS
An AI agent just deleted your data. Do you know why, who authorized it, what path it took—and how to stop it before it does more damage?
As AI agents move into engineering workflows, employee experiences, and enterprise operations, they can no longer be treated like standard service accounts or API keys. Agents need to be treated as first-class identities with owners, lineage, authorization, and a kill switch when risks escalate.
In this special edition of the Okta Streamcast, Okta CEO and Co-Founder Todd McKinnon joins Harish Peri (SVP and GM for AI Security) to reveal the blueprint for the...
Questions asked in this interview
7
0:05Today we are joined by our own Octa CEO and co-founder Todd Mckinnon. Todd, how you doing?
Hey everybody, welcome back to the latest episode of the Octa Dreamcast. Today we are joined by our own Octa CEO and co-founder Todd Mckinnon. Todd, how you doing?
Fantastic. So today's format is a little different. We'll hear from Todd about his point of view on zero trust security and what organizations should consider before rolling out agents. After that, we'll go into live demos of exciting new innovations. Let's jump in. Todd, a couple months ago, we launched the blueprint for the secure enterprise. Why a blueprint? What's in it?
I think a lot of what we do is informed by conversations with customers. They're excited about AI agentic capabilities but confused about how the pieces fit together. One customer said we need a plan for how it all fits. So we put together the blueprint to explain to customers and prospects how this world could fit together, what role Octa plays, and what Octa isn't doing. We have cool capabilities, but where's the borderline and what other technology should they put around Octa to manage and secure this next generation?
Yeah, that makes sense. So when an organization implements the blueprint and has that architecture rolled out, what does that look like? How do they know they're there?
They get trust and confidence that this new layer of technology is providing visibility, control, and risk mitigation while delivering business value. It's about driving the business, automating, reducing costs, and opening new revenue avenues, but managing and securing it all so they don't open up their risk aperture.
Yeah. I've been hearing more about zero trust. Identity is core for humans, but with agents, what's your POV on how zero trust needs to evolve in the age of agentic?
And you would say it doesn't have to? I'm just curious how you would position it.
I think about it in two layers. Zero trust was a new concept 15 years ago, with the BeyondCorp paper from Google. The term was coined by Gartner, and it set a stake in the ground for where the industry was headed. Similarly, we need a zero trust for agents. Specifically, we need the equivalent of phishing-resistant MFA for agents. Token theft is a big issue with credentials in source code. At a high level, we need a zero trust architecture for agents, and tactically we need phishing-resistant MFA for agents now.
So along those lines, the point of this screencast is to get into practical demos. Many customers are looking to roll out agents into production. If you were to give a CISO or CEO three pro tips for rolling out agents – things like identity, kill switch – what would those be?
First, think about agents as a new version of everything in the stack. Agentic is a broad transformation across cloud providers, infrastructure, applications—all layers become agentic. So you need visibility across the whole stack to see where agents are coming from. Second, focus on connections: what agents are connecting to. More connections make agents more powerful, but you need secure authentication for those connections. Third, define what agents can do downstream—the authorization so they only act as intended based on their task. Those are the three things: catalog agents across the stack, control connections, and control actions.
Agents are a new version of everything in the tech stack. Just like a kill switch is valuable in security for zero days and breaches, you want the ability to isolate and cut off the agentic layer when new attack patterns emerge.
13 more exchanges in this transcript
Sign in free to read the rest of this interview. No card required.
Mckinnon, T. (2026, June 18). How to Discover, Connect & Govern AI Agents | Okta CEO Todd McKinnon & SVP AI Security Harish Peri [Interview transcript]. Okta. CEOInterviews.AI. https://ceointerviews.ai/interview/1052988/
MLA
Todd Mckinnon. "How to Discover, Connect & Govern AI Agents | Okta CEO Todd McKinnon & SVP AI Security Harish Peri." Okta, 18 Jun. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/1052988/.
BibTeX
@misc{mckinnon2026_1052988,
author = {Todd Mckinnon},
title = {How to Discover, Connect \& Govern AI Agents | Okta CEO Todd McKinnon \& SVP AI Security Harish Peri},
howpublished = {Interview transcript, Okta. CEOInterviews.AI},
year = {2026},
month = {jun},
url = {https://ceointerviews.ai/interview/1052988/},
note = {Speaker-attributed transcript with timestamps}
}
