€825 million. That’s the price for letting an algorithm make decisions without any human involvement.

Last week, former FTC chair Lina Khan called AI self-regulation a “proven failure.” She recommends focusing on enforcing the laws that already exist.

That’s already happening in the Netherlands. In August, the Dutch Data Protection Authority fined Uber nearly €825 million.

According to the Dutch DPA, Uber’s systems blocked driver accounts between 2018 and 2022 over suspected fraud or low ratings, without any human review. Drivers lost their income overnight. Uber is appealing.

No AI Act was involved. This was plain Article 22 GDPR, which prohibits fully automated decisions with significant effects on individuals, such as loss of income.

Now imagine replacing “driver” with “customer” and “low ratings” with “high risks”. Fintechs and banks have always relied on automation. Automated models often trigger account freezes, stop payments, and client removals. I’ve contributed to automated controls in banks and fintech myself, so I understand how tempting it is to let the system make the call.

So the real question isn’t whether there’s a human in the loop. It’s whether that human actually counts.

The Dutch DPA is clear about this. You need someone who truly reviews each case and can change the outcome. Just clicking “approve” isn’t enough. This isn’t just a technology issue.

It’s about governance:
→ Who owns the model?
→ Who can override it?
→ How is that documented?
→ Does the customer know where they stand?

I always say: you can’t stop a wave, but you can learn to surf it. AI is that wave, and waiting until the AI Act’s high-risk rules kick in in December 2027 isn’t a strategy. Are you paddling out, or still watching from the beach?