CrowdStrike builds an identity provider for AI agents, not humans

CrowdStrike is drawing a hard line between human logins and machine ones. At Fal.Con, the company unveiled an Agentic Identity Provider built for AI agents, which it says need to be identified before anyone can decide what they’re allowed to do.

That matters because most companies are still faking agent identity with service accounts and API keys. CrowdStrike’s new provider sits ahead of its Continuous Identity system, which the company introduced in June using technology from its $740 million SGNL acquisition. Continuous Identity handles authorization in real time. The new piece is about the earlier question: what, exactly, is this agent?

Jennifer “JJ” Johnson, CrowdStrike’s chief marketing officer, framed the whole thing around a blunt ratio: roughly 90 AI agents for every human employee. Her point was simple. Security teams can’t keep treating agents like users with faces, names and a neat chain of ownership. They’re different by design, and the old model was never built for that.

CrowdStrike is also pairing the identity push with what it calls the next step in its Agentic SOC. The idea is coordinated investigations that follow an attack across endpoint, identity, SaaS, cloud and network at the same time. Johnson’s argument was that defending AI with AI means seeing across domains and reacting in one motion, not one system at a time. That’s a familiar CrowdStrike pitch, but the agent piece gives it a sharper target.

One CISO, Johnson said, admitted that the industry never really solved human identity in the first place. Her spin was almost cheeky: maybe agents are the chance to get it right, and work backward from there. That’s a big promise. It also sounds exactly like the kind of problem that gets worse the moment companies decide they can probably wing it.