Block

CA501-Agents-IdentityProtection-AnyApp-AnyPlatform-BLOCK-HighRiskAgent

Block access

About this policy

The CA501-Agents-IdentityProtection-AnyApp-AnyPlatform-BLOCK-HighRiskAgent policy blocks agent identities whose identity risk is rated high. As AI agents take on real work in Microsoft 365, they become identities worth stealing, and a compromised agent can act at machine speed. This policy covers all agent users across every resource, cutting access the moment risk signals flag that something is wrong. Blocking rather than challenging makes sense here, since an agent cannot exactly complete an MFA prompt to prove its innocence.

Policy conditions

Users: All agent users
Apps: All cloud apps
Risk: Agent identity risk: high

JSON template

{
  "@odata.context": "https://graph.microsoft.com/beta/$metadata#identity/conditionalAccess/policies/$entity",
  "templateId": null,
  "displayName": "CA501-Agents-IdentityProtection-AnyApp-AnyPlatform-BLOCK-HighRiskAgent",
  "modifiedDateTime": null,
  "state": "enabledForReportingButNotEnforced",
  "deletedDateTime": null,
  "partialEnablementStrategy": null,
  "sessionControls": null,
  "conditions": {
    "userRiskLevels": [],
    "signInRiskLevels": [],
    "clientAppTypes": [
      "all"
    ],
    "agentIdRiskLevels": "high",
    "platforms": null,
    "locations": null,
    "times": null,
    "deviceStates": null,
    "devices": null,
    "clientApplications": null,
    "applications": {
      "includeApplications": [
        "All"
      ],
      "excludeApplications": [],
      "includeUserActions": [],
      "includeAuthenticationContextClassReferences": [],
      "applicationFilter": null
    },
    "users": {
      "includeUsers": [
        "None"
      ],
      "excludeUsers": [],
      "includeGroups": [],
      "excludeGroups": [],
      "includeRoles": [],
      "excludeRoles": [],
      "includeGuestsOrExternalUsers": null,
      "excludeGuestsOrExternalUsers": null
    },
    "agents": {
      "includeAgentUsers": [
        "All"
      ],
      "excludeAgentUsers": [],
      "agentFilter": null
    }
  },
  "grantControls": {
    "operator": "OR",
    "builtInControls": [
      "block"
    ],
    "customAuthenticationFactors": [],
    "termsOfUse": [],
    "authenticationStrength@odata.context": "https://graph.microsoft.com/beta/$metadata#identity/conditionalAccess/policies('55b1698b-889c-4420-b6f0-85ed0df5f4a4')/grantControls/authenticationStrength/$entity",
    "authenticationStrength": null
  },
  "summary": "The CA501-Agents-IdentityProtection-AnyApp-AnyPlatform-BLOCK-HighRiskAgent policy blocks agent identities whose identity risk is rated high. As AI agents take on real work in Microsoft 365, they become identities worth stealing, and a compromised agent can act at machine speed. This policy covers all agent users across every resource, cutting access the moment risk signals flag that something is wrong. Blocking rather than challenging makes sense here, since an agent cannot exactly complete an MFA prompt to prove its innocence."
}