Qos|cWhere AI is used
The chat assistant, document translation, rewriting and plain-language editing, summaries, meeting analysis, transcription of lesson audio, and meaning-based search. No other feature calls an AI model.
What is sent, and what is not
Only what the task needs: your question, the content of the record you are working on, and that record’s context inside your organization. Identity numbers, IBANs and card numbers are masked before the request leaves. Requests are built inside your organization’s boundary; another customer’s data never enters the same request.
Who decides
Everything the AI produces is a suggestion. A person approves it before it is saved or published. No fully automated decision with a legal effect on a person is made.
How it is marked
Content produced with AI is marked on the record: which task ran, when, and who accepted it. If a person edits it afterwards, that shows too. So where a text came from can still be read later.
Your right to turn it off
Every user can turn AI off in their own profile. While it is off the system runs no AI operation on that person’s behalf; the request is refused on the server. An administrator can also manage this across the organization.
Model training
Your data is not used to train AI models, and the agreement with the provider states the same. We do not train a customer-specific model on your data.
Records and audit
Every AI call is recorded: which organization, which user, which task, when, and how much usage. The content itself is not kept in that record. An administrator can see their own organization’s usage summary in the product.