AIBRIEF
The three vendors each cover a different layer of agent risk, feeding behavioral, data, and app level signals into Microsoft's new Agent 365 control plane for AI agents.
An AI agent can draft a customer email, file a support ticket, and move money between systems without a human in the loop. Microsoft has picked three security vendors to watch those agents inside Agent 365, the new control plane the company is building for software actors that act on a user's behalf.
The inaugural cohort, named in the July 2026 Agent 365 update, is Darktrace, Cyera, and Zenity. Each covers a different layer of agent risk. Darktrace contributes behavioral signals drawn from a per-customer baseline of normal agent activity, delivered through its / SECURE AI product. Cyera focuses on the data agents can reach. Zenity covers the apps and actions agents perform. That split shows how Microsoft wants the agent-risk surface partitioned.
Ed Jennings, President and CEO of Darktrace, said agent risk visibility is critical as agents take on more responsibility across sensitive data and systems (iTWire coverage).
The integration is a stated capability, not a deployed one. The announcements name no customer rollout, no measurable outcome, and no general availability date. Microsoft is locking in the vendor mix before most enterprises ship agent deployments at scale, which is also when category framing gets cemented.