Manage AI agent permissions

AI Agents Permissions lets you control which activities AI agents are allowed to perform on your applications. You can tailor policies for individual agents, which gives you full control over their permissions based on their unique characteristics and your organization’s needs.

📘

Note

You can only manage permissions for AI agents that are known to HUMAN. If there’s an agent that isn’t included here that you believe should be, then we encourage you to Suggest a new AI Agent.

You can learn how to manage AI agent permissions with this article. For information on managing policies for AI crawlers, see Traffic Policy Settings.

Prerequisites

You need certain Role permissions to update AI agent policy settings. By default, this role is an Admin in Sightline.

Update permissions

To change permissions:

1. Navigate to Sightline Cyberfraud Defense > Policies > AI Agents Permissions.
2. Select the Application you want to apply the permissions to. You can set different permissions for each application.
3. If needed, search for the AI agent you want to update.
4. Toggle each individual permission for that agent to allow or deny as needed. Alternatively, if you want to deny all permissions, you can toggle on a checkmark in the Deny column.
5. Set a rate limit for the agent as well. This limits the agent’s allowed requests per minute during its session.
6. Click Save changes.

Your AI agent permissions are now updated.

About the Permissions table

The Permissions table has the following information:

• Agent Name, Provider, and Description: The name of the agent, the agent’s provider, and a description.
• AI Agent Type: The type of agent it is.
• Trust: How trustworthy AgenticTrust considers this agent. AgenticTrust calculates this based on how verified the agent is as well as the spoofability of the agent. AgenticTrust measures Trust on three levels:
  • High (H): The agent is fully verified and difficult to spoof.
  • Medium (M): The agent is semi-verified and partially declared.
  • Low (L): The agent is unverified and easily spoofable.
• Volume (7d) (Requests): The total number of requests from the agent over the past seven days.
• Permissions: The available permissions you can allow or deny for each agent. Each permission is described below. See Agentic activity priority for detailed information on which activities are associated with each one.
  • Deny: Block all permissions for this agent on your application.
  • Read: Permission for the agent to read data on your site. This is the most basic and default permission if you don’t deny an AI agent. However, if this is the only permission you allow, then AgenticTrust will block the agent from navigating to or reading pages related to the remaining permissions such as login or checkout pages. If you want to allow the AI agent to conduct checkouts or log in, you’ll need to toggle on the corresponding checkbox to enable the permission.
  • Engage: Permission for the agent to engage with users and content on your application such as sending messages, commenting, playing media, and so on.
  • Account Creation: Permission for the agent to create new accounts.
  • Login: Permission for the agent to log in on a user’s behalf.
  • Change Account: Permission for the agent to make changes to a user’s account details such as their email address, password, phone number, credit card information, and so on.
  • Checkout: Permission for the agent to purchase items from your site, including as a guest user.
  • Rate Limit: The number of allowed requests the agent can make per minute during a session.

Default permissions

If you don’t set your own permissions, HUMAN will assign default permissions for each known AI agent. HUMAN bases these defaults on the agent’s individual level of trust and the types of activities it tends to perform.

Default allowed permissions are denoted by a star . Meanwhile, any permissions you set do not have a star, as shown below.

• : You’ve set the permission as allowed.
• : You’ve set the permission as denied.

If you set a permission yourself, then those permissions will persist regardless of HUMAN’s defaults. This means that, even if HUMAN changes its default permissions later, yours will not change and will continue to be whatever you originally chose.