When the EU Regulates, the Classroom Shifts: Firsthand Lessons from Teaching AI Skills in Germany

Sunday Reflection — 2026-06-18

First week of the new AI curriculum block in Berlin. One student puts her hand up: "Dean, if the EU bans this model, do we just bin the project?" Nobody in the room cared about AI hype. They wanted to know what keeps the company out of trouble. That is the shift. Six months ago AI ethics felt like box-ticking. Now they want to know which tools survive an audit and which ones land someone in court.

Last Tuesday I had to rewrite half a lesson overnight. The EU dropped fresh guidance on model explainability. The German approach arrived in full: process maps, checklists, a worksheet that reads like a driving theory exam. These students can define red teaming and GDPR obligations inside out. But half of them could not explain to a product owner why you cannot feed customer data into a US-based API without checking the data processing agreement first.

When I guest-lectured at a UK training partner last autumn the contrast was stark. Less paperwork. Much more "what works, how fast can we try it, we will fix it after." The Germans wanted to pass the audit. The Brits wanted something to ship next week.

flowchart LR
  A[New EU guidance drops] --> B{Team response}
  B -- Germany --> C[Update compliance docs]
  C --> D[Check if tool still approved]
  B -- UK --> E[Prototype first]
  E --> F[Check compliance when challenged]
  D --> G[Deploy carefully]
  F --> G

Neither is wrong. My first experience of this gap was in post-unification Germany: laying field cables while our German counterpart waited for written orders confirming the route. We wanted to finish before dark. They wanted everyone standing in the right place when it happened. That tension has not gone away. AI regulation has just sharpened it.

The big lesson this week: regulation changes what students need to know, but it does not teach anyone which rules actually matter or how to handle the guidance that drops on a Thursday night.