Okta calls for shared AI agent security standards

Okta is urging the industry to adopt coordinated safeguards for AI agents, framing the challenge as a shared responsibility rather than one vendors can address in isolation. The call comes as autonomous systems increasingly move across platforms, creating security gaps that individual controls can’t close. Charlotte Wylie, Okta’s senior vice president and deputy chief security officer, has suggested that trust in AI agents may require broader alignment beyond technical measures alone.

This isn’t just a theoretical concern. Over the past month, StartupReader has tracked developments suggesting that AI agent risks are already taking shape. Island’s recent $400 million round, at a $6.4 billion valuation, appeared to signal growing investor confidence in securing these systems—particularly within enterprise browsers. Palo Alto Networks, for its part, has indicated it is adjusting its security approach to account for agents that operate with greater autonomy. Meanwhile, reports have circulated about a possible government breach involving AI agents accessing systems beyond their intended scope, though details remain unconfirmed.

The tension here is clear: AI agents appear to be evolving faster than the governance tools meant to contain them. The rapid adoption of these systems—highlighted in recent *theCUBE* discussions—has outpaced the development of standardized safeguards, leaving enterprises to navigate a patchwork of vendor-specific controls. Okta’s position suggests that this approach may not be sufficient. If agents can operate across applications and platforms, security measures confined to a single product or company’s ecosystem may struggle to address the full scope of the risk.

What’s notable about Okta’s stance is its implicit critique of the industry’s current trajectory. While some startups and incumbents are focused on building proprietary solutions, Okta is arguing that the problem may require collaboration. This aligns with a broader trend StartupReader has observed: as AI agents become more capable, their risks—such as data leakage or unauthorized actions—could extend beyond individual companies’ boundaries. A security gap in one system might have implications for others, making isolated measures potentially ineffective.

The question now is whether Okta’s call will gain traction. The company hasn’t proposed a specific framework, and without regulatory pressure or a clear economic incentive, vendors may be reluctant to adopt shared standards. But the alternative—continuing to treat AI agent security as a product-level problem—could prove increasingly difficult to sustain. If further incidents emerge, they might force the industry to reconsider its approach, much like past breaches have driven shifts in cloud security and data privacy.

For now, the conversation remains fragmented. Some companies are investing in technical solutions, others are waiting for standards to emerge, and a few, like Okta, are advocating for a more coordinated approach. The coming months may reveal whether the industry can align on shared safeguards—or whether AI agent security will remain a challenge left to individual players.

Sources: siliconangle.com