To investigate an AI agent’s activity, check the connected account and the affected service’s own history or recovery tools. To stop further access, disconnect the app or revoke the permissions you control. These are separate steps: disconnecting an agent does not automatically reverse changes it already made.
First identify the agent, account, and time
Write down the agent’s name, which provider account it used, where you noticed a change, and the approximate time. An agent’s reach depends on the connection and permissions granted to it, plus any controls imposed by a workplace administrator; it does not automatically have access to every account you own. Review the actual account and permission grant rather than relying on the agent’s name or intended task.
As an Amazon Associate I earn from qualifying purchases.
For ChatGPT-connected apps, OpenAI documents a connected-account view at Settings > Plugins. Select the relevant app to review its connection. The available accounts and actions depend on the app, provider account, workspace controls, and availability for your plan, region, account, and product surface. See OpenAI’s connected-app guidance.
Free tools Windows power users keep installed
One-click scans. No signup required.
For a managed Microsoft Agent ID deployment, an organization can inspect the agent identity, its granted permissions, and sign-in logs in Microsoft Entra. These are Microsoft-specific administrative controls, not general steps for consumer AI apps. Microsoft’s administrator documentation explains the available identity and audit views, and notes that viewing or managing them requires the appropriate account, role, and, for some controls, licensing: Microsoft Entra Agent ID documentation.
Find out what changed
Start with the service where the change appeared, because a provider’s account or sign-in history may not show every action the agent performed inside another service. Depending on the type of change, look for that service’s activity or edit history, sent items, trash or recovery area, version history, transaction history, or administrator audit records. Use the service’s own documented recovery process; there is no universal AI-agent rollback control established across providers.
Keep relevant evidence before changing settings if the event may need investigation: note timestamps, the agent name, the permission grant, and any relevant activity or sign-in records. Microsoft’s end-user guidance for suspicious Agent ID activity says: “Capture a screenshot, note the agent name, and contact your helpdesk or security team for guidance.” See Microsoft’s Agent ID sign-in guidance.
Sign-in logs can help establish which identity authenticated and when; they may not enumerate every downstream edit, message, or other action in a separate service. For a work or school account, ask the administrator to check relevant audit records if your own service history does not explain what happened.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Stop the agent from accessing the account again
ChatGPT-connected app
- Open Settings > Plugins in ChatGPT.
- Select the connected app and open its account or connection controls.
- Choose Disconnect for the account you want to unlink.
OpenAI says disconnecting stops future access through that account. It does not automatically delete existing or archived conversations or saved memories, and it does not itself undo earlier changes. Other connected accounts or administrator-managed connections may remain active. See OpenAI’s connected-app guidance for the product’s connection controls.
Microsoft work or school app
- Open the app in the My Apps portal.
- Choose Manage your application.
- Review the permissions you personally consented to, then choose Revoke Permissions if you want to remove them.
Microsoft warns that revoking permissions can break some app functionality. My Apps distinguishes permissions you consented to from those consented to by an administrator; you cannot revoke administrator-consented permissions there. Contact your work or school administrator about those grants. See Microsoft’s My Apps end-user guidance.
Microsoft Entra Agent ID deployment
For an organization-managed Agent ID, an authorized administrator or agent owner can use Entra’s identity controls to review or manage the agent identity and permissions. Disabling an identity or changing its permissions is an administrative action, and the available controls depend on the person’s role and the organization’s licensing. If you are a consumer or ordinary end user, contact your organization’s administrator rather than attempting tenant-wide changes. Microsoft documents these controls in its Entra Agent ID guidance.
Understand which control applies
| Situation | Who can act | What the control does |
|---|---|---|
| Personal connection you authorized | You can review and disconnect that account in the connected app’s settings. | Stops future access through that connection; it does not reverse completed actions. |
| Work or school permission you personally consented to | You can review and revoke it in My Apps. | Removes that permission; some app functionality may stop working. |
| Permission or connection managed by an administrator | Your workplace or school administrator must review or change it. | The user cannot revoke administrator-consented permissions through My Apps. |
| A completed edit, deletion, message, or transaction | Use the affected service’s own history, recovery controls, or support process. | May help restore or investigate the change if that service offers an applicable option; disconnecting alone does not do so. |
| Microsoft Entra Agent ID identity or permissions | An authorized agent owner or administrator, subject to role and licensing. | Can manage the organization’s agent identity and permissions; it is not a consumer-app control. |
Make future approvals more deliberate
Connection authorization and action approval are different layers. A connection determines what account access the app has; an action-approval preference determines when the assistant asks before reading or doing something. OpenAI documents permission preferences for when ChatGPT asks before reading or taking an action. Changing that preference does not grant new provider access; removing access requires disconnecting the app or asking an administrator to disable a managed connection. Where a product offers the choice, use prompts for consequential actions and grant only the account and scope needed for the task. See OpenAI’s connected-app guidance.
For Microsoft Agent ID, the consent prompt identifies the agent or publisher and the requested data or actions. Microsoft explains: “It doesn’t give the agent unlimited access.” Approval allows use of the listed permissions without asking each time and may permit background operation for configured tasks; additional permissions require another grant. Check the publisher and each requested permission before approving. See Microsoft’s Agent ID sign-in guidance.
When to involve an administrator or provider
- The connection or permission is labeled as administrator-managed, or you cannot revoke it yourself.
- The agent may have made a high-impact change, such as altering important work records or sending sensitive material.
- The affected service’s history does not explain the change, or you cannot tell which account or identity was used.
- You suspect suspicious activity and need help preserving or interpreting organization logs.
For a workplace incident, share the agent name, account, approximate time, screenshots, and relevant records with your helpdesk or security team. Avoid deleting evidence while the activity is being investigated.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
