AI and your data
Which features use AI, exactly what each one sends, and what never leaves. Written by reading every place in the product that calls an AI model.
Last updated 17 October 2026
This is the complete list. There are no other AI calls in the product.
Anthropic, a United States company, processes each request to produce the answer. Under its commercial terms it does not use that content to train its models. It may keep requests and answers for a limited time for safety and abuse monitoring, as its privacy policy describes.
Separately from the features above, a staff member can connect their own AI assistant (for example Claude Desktop) to Farband with a token they create. That assistant can never do more than the person who created the token, can be limited to less, and every action it takes is recorded. In that case the data goes to whichever AI service the staff member chose, under their organization’s own account and terms, not ours.
Each organization has one switch, Use AI features, that its owners and admins can change (Admin, Branding & naming). It is on by default. Turned off, nothing about that organization is sent to Anthropic from anywhere in Farband: every feature in the table above stops, and the help assistant with it. The check runs on our servers on every request, so it holds even for a page that was already open when the switch changed. The buttons for those features hide or explain that AI is off, and everything you can do by hand keeps working.
The switch is per organization: turning it off in one workspace changes nothing for any other. Every change to it is recorded in that organization’s audit log. It does not cover an AI assistant a staff member connects themselves (above), since that data goes to the service they chose; an admin can revoke those tokens separately.
Questions
Privacy questions are welcome, including the awkward ones. Write to hello@farband.co.