The EU AI Act in the Classroom and Codebase: A Practitioner’s View from Saxony-Anhalt
Sunday Reflection — 2026-09-13
Out of nowhere, half my slides turned into legal checklists. That’s how the EU AI Act arrived in my classroom. One week we’re talking about transfer learning, the next I’m warning students which models now need a data provenance audit. People keep saying the Act is still theory, but it’s already in my inbox and in my teaching. Students want clear rules. The law delivers fog.
The change isn’t clean. At Masterschool in Berlin, we replaced two class projects because there’s no time to untangle if Kaggle datasets tick all the boxes. I stop every demo and ask: “Where did this data come from? Who consented?” That’s new. Meanwhile, with consulting clients, I’ve spent days rewriting project documentation. Small German firms have gone from barely mentioning data handling to thick binders signed by every product owner. More paperwork means less actual coding, at least in the short run. The joke about German process obsession used to bug me. Now I see the point: when the rules really come, you’d rather have your paperwork lined up than improvise with a British shrug.
mermaid flowchart LR A[EU AI Act Begins] --> B{Are Data Sources Documented?} B -- Yes --> C[Proceed with Project] B -- No --> D[Pause Course/Project] D --> E[Source & Document Data] E --> B
This feels a lot like my days building bridge layouts in the Army. We drew up stacks of plans not because the bridge would follow them, but so you could point and say: we thought this through. Same now for code. Most British teams I know would still rather risk it and fix on the fly, then moan when audit season comes. German teams try to pin every detail up front. Both have blind spots. The British style misses the hidden legal tripwires. The German one risks paralysis by paperwork. It isn’t about finding the perfect approach. I’m watching both groups stumble into the same confusion, but in different accents.
If you’re teaching or building in Germany, the Act isn’t a future headache. It’s your Monday morning. One new rule: don’t expect clarity—just make your best record at every step and accept that it’ll probably have to change.