Walk the show floor at Oktane 2026, and it’s clear: AI agents are everywhere.
From writing code and drafting emails to analyzing complex datasets, AI tools have rapidly moved from consumer novelty to enterprise reality. Organizations worldwide are under pressure to deploy AI swiftly without sacrificing security.
“The rapid adoption of AI and agentic identity is something that every enterprise is trying to deal with,” said Charlotte Wylie, SVP and Deputy CSO at Okta.
Oktane sessions highlighted a central challenge to that adoption: Agents need access to deliver real value, but that same access creates security risks. The answer, according to Okta CEO and Co-Founder Todd McKinnon, isn't to lock them down or let them run free—it's visibility and control.
The blueprint for the secure agentic enterprise
In his keynote address, McKinnon introduced the blueprint for the secure agentic enterprise, a reference architecture to help organizations maintain visibility and control as they deploy agents at scale.
The Blueprint Alliance, the industry coalition behind the blueprint, is committed to setting standards, developing integrations, and exchanging risk information to help ensure a more secure agentic future, according to Matt Ellard, SVP and General Manager for EMEA.
“If we have a rich ecosystem of integrations across the [alliance] members, that means we're exchanging threat signals, federating our agent registries with each other, and using each other's data to facilitate agent import and stop shadow AI,” he explained.
The risks at hand
On the ground, attendees and enterprise leaders shared concerns over emerging AI risks. As organizations adopt new AI tools, the risk of data leaks and shadow AI—deployments operating outside established security frameworks—grows sharply.
“The biggest threat is shadow AI,” Ellard said. “In those situations, users aren't adhering to a framework that ensures security, provides visibility into AI activity, and promotes ethical and responsible AI usage.”
And when autonomous agents connect directly to sensitive enterprise data and core systems, traditional perimeter defenses aren't enough.
Treating AI agents as first-class identities
This is where identity comes in. Enterprises have long turned to identity providers to manage human identities and access. Now the industry is looking to apply those same controls and standards to AI agents.
In response to this growing demand, Okta introduced new innovations to Okta for AI Agents as well as Agent SSO to help organizations maintain visibility and control over their agents.
As Okta partners and customers observed throughout the week, navigating these uncharted waters isn't about slowing down adoption. It's about striking the balance between security and innovation.
Stay tuned for more Oktane coverage on how companies are finding their own balance, navigating new AI regulation, and preparing for emerging AI risks.
