Okta's Wylie: agent security needs shared safeguards

Okta is framing agentic AI security as a problem that no single vendor can close on its own. Its Blueprint Alliance, built with founding members including Amazon Web Services, CrowdStrike, Salesforce and ServiceNow, aims to make security controls and risk signals work across different agentic systems. The pitch is less a product launch than a shared architecture: Charlotte Wylie, Okta's senior vice president and deputy chief security officer, described an approach where each agent is given an identity, has its permissions limited, and is monitored for how it behaves.

Wylie made the case at Okta's Oktane event in a conversation with theCUBE Research's Krista Case and co-host Rebecca Knight. Her argument is that customers are already scrambling to adopt autonomous agents faster than any one vendor can secure them, so the industry has to galvanize rather than solve the problem alone. That is where Okta for AI Agents fits into the picture she described, alongside the governance and visibility controls that let an organization stop an agent that goes rogue.

The less technical part of Wylie's message is that this has to be explained outside security teams. She said she hopes the messaging about how Okta is steering into this carries well beyond the conference and its attendees, so that people go home to their families and feel they understand how it will be done. Whether that lands may depend on whether the alliance's controls are actually interoperable in practice, and on whether the safeguards it describes reach the business people who will end up running these agents.