Prompting ohne schlechtes Gewissen

A talk at the MAI-Tagung 2026 (Historisches Museum Frankfurt, 28 May 2026), together with Christoph Wetzel, in the opening block KI-Impulse für Museen. Subtitle: Wie das ZKM seinen Mitarbeitenden sichere KI-Superkräfte gab (how ZKM gave its staff safe AI superpowers).

What the talk is about

It starts from three scenes every museum knows: an educator uploading books to a US provider, a curator letting ChatGPT write an object text, a trainee who doesn’t know what she is allowed to do. Staff wanted to use AI, and they should not have to do it in secret. “Without a guilty conscience” means legally, technically, financially and ecologically.

ZKM’s answer is ask.zkm.ai, one portal for everyone, built on four guarantees:

• Identity. Login only with the ZKM account, with no second password and no anonymous access.
• Sovereignty. Academic models hosted in Germany as the default; US providers only where needed, and visibly marked.
• Costs. Transparent across all models. Who uses which service how much is visible before the invoice arrives.
• CO₂. Consumption is shown in the chat, and publicly, in real time.

Then it covers what staff do with it today (object texts and alt texts, talking to the collection, image generation for exhibition graphics, safe web research, working through long documents) and what other houses can take away: one door to the outside for every model, and a human in the loop.

The technical deep dive into the same portal followed in October at kulturBdigital: Prompting Portal.