Analysis
Okta agreed to acquire Permiso, an AI identity security startup founded by former FireEye executives Paul Nguyen and Jason Martin, in a deal valued at just under $200 million and structured as almost all-cash. The transaction is expected to close in the third quarter of Okta's fiscal 2027.
Permiso's core technology detects attacks that exploit stolen or compromised identities to move laterally through cloud infrastructure -- a threat category that predates AI agents but has grown more urgent as autonomous agents get granted real system credentials and permissions. In April, Permiso launched SandyClaw, a sandboxed environment purpose-built to analyze AI agent behavior and flag malicious or unintended actions before those agents reach production, technology that becomes central to Okta's platform once the deal closes.
The acquisition fits a broader pattern of identity-security incumbents moving quickly to acquire rather than build agentic-AI-specific capability, mirroring the urgency behind ThreatLocker's own AI-agent security push disclosed the same week. Okta is betting that securing non-human identities -- AI agents, service accounts, machine-to-machine credentials -- becomes as large a market as securing human logins has been for the past decade.
For security-focused investors, the roughly $200 million price on a company that only publicly launched its flagship agent-sandboxing product in April suggests acquirers are paying a premium for teams with proven identity-threat-detection pedigree who've already built working AI-agent security tooling, rather than waiting to build it in-house. What to watch: how quickly Okta integrates SandyClaw into its core platform, and whether the deal price becomes a benchmark for the next wave of AI-agent security acquisitions.
