How we use AI

go~mus is infrastructure for cultural institutions. We approach artificial intelligence in the same way: not as a later addition and not as a chatbot placed in front of the product, but as part of a carefully built foundation. Our task is not to sell museums AI. Our task is to keep their work reliable, accessible and ready for the future, including a world in which software is increasingly read and operated by machines.

1. AI is infrastructure, not a feature

For us, AI is not a later addition. We build go~mus from the ground up to be cleanly structured, documented and machine-readable. That lowers the cost and the time to connect it to a museum's IT and keeps the system ready for the tools to come. No museum makes itself dependent on a single vendor's AI promise this way. A clean, open foundation is what allows AI tools to do useful work at all.

2. The data belongs to the museum

The data belongs to the museum. We do not pass it to AI services and we do not train models with it. That includes our own development: the tools there work on source code, documentation and test data. Production data has no place in a development environment, not even inside a stack trace.

What an institution does with an agent of its own is the institution's decision. Point an agent at our API and the data goes where the institution sends it. We supply documented interfaces and the data-processing agreement. We do not make that call for the house, and we will not claim control over tools that are not ours.

Visitors entrust their data to the museum, and the museum entrusts it to us. And we do not pass it on.

3. For institutions of every size, and we listen

Clean, machine-readable interfaces lower integration costs for the small house with limited IT resources just as much as for the large network. We build for the entire sector, not only for those who can afford extensive consulting. And we listen to the sector. AI does not replace the conversation with the institution, it gives us more time for it.

4. The human decides

We are convinced that AI will be put to good use in many areas, including the museum. Its development, however, has to follow concrete needs in the workflows rather than technology for its own sake. We analyse the workflows and develop tools and processes with AI that help the museum work more efficiently. The machine prepares; the decision is always made by a human.

What we want to be measured by: whether museums can do their work better, not how much AI we claim for ourselves. We work transparently, we stay honest about what the technology can and cannot do, and control remains with the human.

This page states a stance. The legal classification, which role we hold for which system and why the transparency obligations of the AI Act do not apply to go~mus, is set out soberly and separately: go~mus and the AI Act.

Version 2.0, as of 30 July 2026

Version history

• 2.0, 30 July 2026: section 2 split into what we do with the data and what an institution does with an agent of its own. Legal classification under the AI Act moved to its own page.
• 1.1, 1 July 2026: wording sharpened, principle 3 named more clearly.
• 1.0, 30 June 2026: first published.