How to Gain Control of AI Agents and Non-Human Identities

Non-human identities—including service accounts, API tokens, and AI agents—are proliferating in cloud-first environments, often created automatically and left unmanaged with little ownership or oversight, according to The Hacker News. The article notes that AI agents can initiate actions autonomously, frequently with persistent credentials and elevated permissions, and that traditional identity controls struggle to monitor them because they lack conventional signals like device context or user sessions.

Common security gaps include invisibility of NHIs, over-permissioned access, and orphaned identities that linger without owners, creating easy avenues for lateral movement and other threats. The piece advocates a shift to proactive governance through an identity security fabric, emphasising inventory, risk-based privilege management, automated rotation of secrets, and kill switches for autonomous agents, with practical steps such as automated provisioning and deprovisioning.

It highlights Okta’s role in surfacing identities and enforcing guardrails, and mentions a guide on integrating Okta with AWS to manage NHI sprawl. Published on 22 September 2025, the article frames non-human identities as critical access points that require the same governance as human users, according to The Hacker News.