Why the EU-UK AI divergence founders fear isn’t real

Anyone who has ambitions to build a business across Europe will have considered the tax of doing it through 27 separate national systems. EU Inc., as a single optional regime to incorporate once and operate everywhere, would remove a real drag on ambition. 

But for those building an AI company, the fragmentation that decides where you can operate isn’t in company law but in the rules governing the product itself. This has led to a fear from founders that the UK and the EU are drifting apart. One is refusing to write an AI rulebook and the other wrote one that nobody can keep up with, so whatever you build for one market is wrong for the other.

However, if you look past the politics, the two regimes are actually converging, not splitting. The clarity founders are waiting for is closer than they think.

There’s no vacuum to wait out

Founders often want to wait until rules are clear before acting. With AI, there’s little reason to hold back and the market isn’t waiting either. According to the Financial Conduct Authority (FCA), a fifth of UK adults, around 11 million people, already say they’re likely to use agentic AI that acts autonomously within pre-set goals. Consumer appetite is running ahead of institutional readiness.

The UK has declined to write a bespoke AI statute and for now, that appears to be the right choice. A dedicated law would be obsolete within two years given how fast the models move. The EU legislated, then hit the same wall – its Digital Omnibus pushed the AI Act’s high-risk obligations back to the end of 2027 because the technical standards weren’t ready.

For now, neither regime is regulating the technology but rulebook or no rulebook, they are both regulating the outcome, and both hold a named person accountable for it, regardless of what produced the decision – a person or a model.

Strip away the headline difference – one has an AI law, one doesn’t, and the goal stays the same – to show the outcome was fair and name who is answerable to it. 

The FCA’s own Mills Review into AI in financial services, published last week, confirms this. Now is not the time to rush out new rules. The FCA’s outcomes-based approach, built around the Senior Managers regime and Consumer Duty, is “the right basis” for what comes next. The accountability model won’t change. 

What actually decides your exposure

The choice of regime matters far less than the seemingly general anxiety suggests. What decides your exposure to AI regulation isn’t which flag sits on your incorporation papers but whether you can reconstruct what your AI did when a regulator, an auditor, or a complaining customer asks.

Responsible AI is not about ethics in the abstract – the real worry keeping a compliance director up at night is narrower. Their concern is around if they can reproduce a decision in eighteen months, when someone demands to know why it was made. Neural network explainability – opening the model and reading its reasoning back –  is mostly pretence.

What a regulator wants is an evidence trail – which document, which clause, which rule, what the system flagged and who reviewed it. This is the same in both the EU and UK.

That’s not a compliance nicety bolted onto the AI afterwards. It has to be built into the system from the first line of code – observability at every step, not just at the output. A model that gets the right answer 98% of the time but can’t show its working is worth less to a regulated business than one that’s slightly slower but leaves a clean, human-reviewed trail behind every decision it touches. 

The Mills Review makes this point directly: governance is likely to become an enabler of capability, not a constraint on it. Firms that can demonstrate auditability and explainability where needed, as well as test robustly and monitor effectively, will be able to deploy AI more confidently, leaving others behind. That’s the regulator saying, in plain terms, that self-governance is a competitive advantage.

The EU reaches the same conclusion by a different route. The AI Act requires automatic logging, pre-market technical documentation  and human oversight that means something, not a rubber stamp. Like the UK, it supports decisions that can be reconstructed and defended, not just described after the fact.

Build for the destination, not the map

If both regimes are heading for outcomes and auditability, that tells you what to build without waiting for anyone’s permission. There’s no EU Inc. for AI regulation, and there doesn’t need to be. Build systems where every decision can be reproduced and defended, with a human answerable for it.

Do that and you’re compliant in London, Frankfurt, and most places between, because that’s the direction all of them are moving. 

EU Inc. would lift a genuine burden, and Europe should move fast on it but founders waiting for the same tidy fix on AI are waiting for something that isn’t coming, and isn’t needed.

The rules that matter already apply. The work isn’t lobbying for a cleaner rulebook but taking the time to build and govern AI systems that can answer the one question every regulator ultimately asks – what happened, and who’s answerable for it.