Skip to content
AletheiaAletheiaAletheia docs

Enabling and disabling the assistant

The assistant is controlled per organization. It can be turned off entirely, and it stays off until someone turns it back on.

The setting

Section titled “The setting”

An organization administrator sets it from the organization settings. Turning it off takes effect immediately:

  • The assistant is removed from the product for everyone in the organization.
  • Requests to the AI endpoints are refused with ai_disabled.
  • Existing conversations and the audit log remain readable. Turning the feature off does not erase the record of what was asked while it was on.

Who can change it

Section titled “Who can change it”

Changing the setting needs the organization-management permission — the same one that covers other organization-wide settings. Being able to use the assistant is ordinary read access; being able to enable it is not. See Roles and permissions.

The change is recorded in the audit log like any other administrative change.

What being enabled does and does not grant

Section titled “What being enabled does and does not grant”

Enabling the assistant grants no new access to anything. Every query it runs re-checks the asking user’s permissions, so two people asking the same question can get different answers, and someone who cannot read a resource cannot read it through the assistant either.

It is a different interface to data people already have, not a wider one.

What leaves your organization

Section titled “What leaves your organization”

Answering a question sends the question and the retrieved records to the model provider. Nothing else in your infrastructure is sent, and nothing is sent for an organization with the assistant disabled.

If that is not acceptable to your security review, leaving the assistant off is a supported configuration — the rest of the product does not depend on it.

Where to go next

Section titled “Where to go next”