What Zenity announced
Zenity, which sells an AI security and governance platform for agents, said it had expanded its coverage across the Microsoft ecosystem. The named surfaces are GitHub Copilot, Copilot Studio, Agent Builder in Microsoft 365 Copilot, Microsoft Copilot Cowork, Microsoft Foundry, and Claude Enterprise in Microsoft 365. The company also described "high-fidelity" alerts streaming into Microsoft Sentinel.
The release does not treat these surfaces equally. The depth of coverage differs by product, as the table shows.
| Surface | What Zenity says it provides |
|---|---|
| Claude Enterprise in Microsoft 365 apps | Telemetry, findings, and inline prevention |
| GitHub Copilot coding agent (GitHub Enterprise) | Telemetry, findings, and inline prevention; Entra ID is used to resolve actor names that enterprise settings hide |
| Copilot Studio (Microsoft Copilot harness) | Telemetry and limited inline prevention |
| Agent Builder in Microsoft 365 Copilot | Telemetry and findings |
| Copilot Cowork | Telemetry and findings, based on newly available Cowork transcripts |
| Microsoft Foundry | Integration migrated to the new Foundry Agent Service |
The Agent Builder, Cowork and Foundry items are described in terms of telemetry, findings or migration. Inline prevention is not claimed for Agent Builder or Cowork.
The Sentinel and "dynamic agent risk" piece
The release says the upgraded integration replaces raw-log export with alerts streamed into Sentinel. Analysts can then triage agent issues in the same queue as other incidents. The initial integration covers supported Copilot Studio and Foundry agents. It does not cover the full list of six surfaces.
For supported agents, eligible Zenity findings can feed Microsoft-calculated dynamic agent risk, and Zenity stays attributed as the source. Microsoft combines eligible runtime alerts from itself and partners with posture signals, including exposure and blast radius, to produce one risk level per agent.
Microsoft's own Ignite material backs this up and adds detail:
- Dynamic Agent Risk is in preview.
- Microsoft's Book of News says a partnership with Check Point, Cyera, Darktrace and Zenity brings partner agent-security signals into Sentinel as evidence for Agent 365 risk.
- The risk level appears in the Agent 365 registry in the Microsoft 365 admin center. The same level is shown in Defender, Entra, and Purview.
Zenity's release hedges on timing, saying the assessment will appear in those portals "as supported capabilities become available." Microsoft's preview label is the more reliable guide to maturity. Treat this as a preview feature, not a finished one.
Context: this builds on earlier work
Zenity and Microsoft had already worked together on Agent 365. Microsoft's July 2026 Agent 365 update said that, when an administrator enables it, Agent 365 can ingest a partner risk signal and match it to the agent through Microsoft Entra Agent ID. Trade coverage at the time said partner integrations are optional and need an admin to enable partner risk signals under Agents settings in the Microsoft 365 admin center. That earlier opt-in model is probably the closest guide to how this will work. The Ignite release itself gives no setup steps.
Zenity's earlier Agent 365 integration covered Copilot Studio, Microsoft 365 Copilot, and Microsoft Foundry. The new announcement therefore extends that work. It is not a first-time partnership.
Why Microsoft shops should care
The release's own scenario is easy to recognize. A coding agent proposes changes in GitHub. A Copilot Studio agent runs a business process. Cowork sends email for the user. Claude builds a forecast in Excel. Each logs activity under its own identity and format.
Microsoft's documentation describes the same problems. Agent 365 extends Defender, Entra and Purview to agents, and lists agent sprawl, over-privileged agents, tool misuse, and prompt injection among the risks. Microsoft's guidance on risky agents says risk signals carry severity and confidence, and deeper investigation continues in existing security tools.
The practical appeal of a partner feed is a single risk level and one queue, instead of separate consoles for each agent surface. The Sentinel routing also means SOC teams do not need to learn a new console for triage.
Caveats
- Vendor claims. Zenity supplied the capability descriptions. The release has no independent tests, customer results, or measurements of detection quality.
- Undefined terms. "High-fidelity," "eligible findings," and "limited inline prevention" are not defined. There are no alert examples, latency figures, or lists of blocked actions.
- Scope gaps. Claude, GitHub, Cowork and Agent Builder coverage is described as visibility and prevention by Zenity. Based on the release, these are not part of the initial Sentinel-to-risk-score integration.
- Missing commercial detail. The release gives no pricing, licensing, tenant prerequisites, or general availability dates.
- Identity resolution. The GitHub claim that Entra ID resolves hidden actor names comes with no stated permissions or failure cases.
What admins should do
- Check which of your agents are on the supported Copilot Studio and Foundry list before assuming they are covered.
- Ask Zenity and Microsoft for rollout status, licensing, and the exact finding types that feed the risk score.
- Review the admin-enablement requirement for partner signals in Agent 365 settings. This is based on Microsoft's earlier guidance, not on this announcement.
- Pilot Sentinel alert volume and tuning before relying on the "one risk level per agent" view for prioritization.
- Keep native Defender, Entra and Purview controls in place. A partner signal adds evidence, and it does not replace those controls.
Bottom line
The announcement is real, and Microsoft has publicly confirmed the Sentinel-to-Agent 365 route. Its value depends on details the release does not provide, so test it in a pilot before committing.
References
- Zenity Expands AI Agent Security Coverage Across the Microsoft Ecosystem - 01net 01net · 2026-11-17T22:35:00+00:00
- Zenity Now Integrates with Microsoft Agent 365 zenity.io
- Darktrace, Cyera And Zenity Among First Security Vendors To Feed Risk Signals Into Microsoft Agent 365 smbtech.au