AI safety controls

Here is what happens to your data when you use the AI assistant.

Conversations are kept for a limited time and then deleted automatically, and nothing you send is used to train a model. This is enforced in the code that runs, not only stated in a policy document.

  1. 01

    You start a conversation with the assistant

    Open the AI Assistant panel from anywhere in the app. It works on the data you give it in that conversation.

    AI Assistant

    Summarise the last 3 emails on this matter
  2. 02

    It is kept for 90 days, then deleted automatically

    A daily sweep deletes anything older than 90 days.

    Retention

    Day 63 of 90, then deleted automatically

  3. 03

    The provider does not train on it

    Each request goes to whichever AI provider is configured for that feature, currently Anthropic and Together AI, or a self hosted Ollama instance a company runs itself. Under contract, third party providers cannot use your data to train their own models, and do not retain it beyond what is needed to return that one response.

    Where a request goes

    Your requestTogether AIAnswer

    Default. Contractually cannot be used for training, and not retained past the response.

    Your requestYour own Ollama serverAnswer

    Optional. Nothing leaves infrastructure you control.

    Two routes, chosen per company