Our commitments

Draft. This is not a fully confirmed general commitment. Commitments 5 to 8 still require confirmation and further detail.

We ask businesses to stay in control of their AI. We need to do the same ourselves. These are the proposed commitments.

1. A person sends what we write to you. An AI colleague prepares drafts. A person reads and sends them.
2. Irreversible actions require a person’s explicit approval first. For example, sending, paying, deleting or publishing.
3. What we build gets a second look before delivery. A fresh reviewer, preferably a different AI model, checks it.
4. We record what you tell us as an agreement, not a quotation. Reports contain goals, agreements and unresolved questions. Direct quotations only if you ask for them.
5. Your data remains yours. We use it only for your assignment. not yet confirmed
6. We do not train AI models on your data. Not on your documents, conversations or figures. not yet confirmed
7. We tell you which AI services see your data. A short overview for each assignment: what runs on our own machines, what runs at a European service and what runs elsewhere. not yet confirmed
8. We share what we learn from our own mistakes. Without client data. Not every mistake, but lessons that can help others. not yet confirmed

Change log: draft (17 October 2026): the first list of eight commitments.