Okta’s AI boom just created a new security problem

Companies worried for years about whether an employee's password had been stolen.

Now they have to worry about AI "workers," too.

Okta (OKTA) reported fiscal second-quarter revenue of $805 million , up 11%, while subscription revenue reached $793 million, up 12%. Remaining performance obligations increased 17% to nearly $4.86 billion .

The company boosted its yearly outlook as organizations increasingly face a category of identity that was virtually nonexistent in mainstream commercial IT only a few years ago: autonomous AI agents.

"Every agent needs a trusted identity and clear controls over what it can access and do," CEO Todd McKinnon said .

For the average worker, the future isn't science fiction. Allowing artificial intelligence to read emails, update customer records, query databases, and trigger corporate processes is becoming more common.

The more power these systems have, the more dangerous a compromised agent can become.

Okta helps govern rising number of digital workers

This issue was a problem businesses were tackling long before autonomous AI took off.

A CyberArk study found that 79% of firms anticipate machine identities growing in the coming year. Almost two-thirds expected growth of up to 50%, with another 16% expecting increases of between 50% and 150%.

AI agents only accelerate that tendency because each autonomous system may need its own authentication credentials, permissions, and audit trail.

If a corporation has an employee in accounting, that company may prohibit them from downloading a complete engineering database.

AI agents should be restricted in the same way.

Earlier this year, Okta released Okta for AI Agents, which is now generally available. The software is designed to find AI agents , enforce least-privilege access, and centrally manage what agents may do.

That creates an entirely new potential security category. Companies may no longer merely pay for Okta to help manage every human employee. They may eventually pay to govern armies of digital workers as well.

The promise of an agent is that users stop doing every step themselves.

An AI agent might read a client complaint, go into the CRM, issue a refund, change inventory, and send a reply. It's super valuable if the agent is allowed to take each step.