Our rules for using AI

By default, all AI models we use in your project run in our own multiply redundant data center in Dortmund – or on-premise at your site. Source code, data models, interfaces and functional specifications never leave our building. Nothing goes to an external AI service without your explicit consent.

For some tasks, models from OpenAI, Anthropic or Mistral are the better choice. We use them if you explicitly ask us to – agreed per project, with a data processing agreement and terms that rule out training on your data. Together, we decide which data may leave the building.

Your source code, your data structures and your functional specifications are not training material – neither for us nor for third parties. We do not train models on client projects, and nothing from your project later flows into other projects. And because we can run local models ourselves at any time, no external provider can make us dependent.

AI doesn't replace engineering – it accelerates it. AI agents deliver a first draft and take on routine work such as test coverage and technical documentation. Our developers steer them and keep writing code themselves so their skills don't erode. Every line goes through code review, static analysis, automated tests and accountable approval by people with a computer science degree. Responsibility stays with us.

On request, we document where and how AI was used in your project – including model, operating location, purpose and review process. Or we don't use it in your project at all. The decision is yours.