0.1%.

That’s UK GDP growth for Q4 2025. Not 1%. Not 0.5%. 0.1%. And into this economy — stagnant, underinvested, productivity-flatlined — the government has decided that what British businesses really need is a new compliance regime for artificial intelligence.

Technology Secretary Peter Kyle told the press that “the British public rightly expect that the most powerful systems are subject to proper oversight.” He’s not wrong about that. But he might be solving the wrong problem.

The economic backdrop

Productivity? Barely moving — 1.1% year-on-year in Q3 2025. Real wages, adjusted for inflation, are barely above where they were in 2008. The UK has had the lowest level of investment among G7 countries for almost every year since 1990.

That’s not a blip. That’s a structural problem. And AI is one of the few technologies with genuine potential to shift the trajectory. The timing of this bill is extraordinary.

The U-turn

Here's what makes this particularly frustrating. Less than three years ago, the UK government published a white paper called "A pro-innovation approach to AI regulation". The entire point was that the UK would not follow the EU down the path of heavy-handed AI legislation. Existing regulators — the FCA, Ofcom, the ICO — would apply principles to AI within their domains. No new laws needed.

That was the pitch. That was why companies were supposed to choose London over Brussels.

Then came the Bletchley Park AI Safety Summit in November 2023 — voluntary commitments from AI developers for pre-release safety testing. Handshakes and press releases. No enforcement.

By November 2024, Peter Kyle was pledging legislation “within a year”. And as recently as January 2026, legal analysts at Osborne Clarke noted that a dedicated AI bill “currently seems unlikely.” The government was focused on AI Growth Zones and Growth Labs instead — regulatory sandboxes designed to attract investment.

Weeks later, Kyle announces the bill.

In under three years, the UK went from “we explicitly won’t legislate” to mandatory registration and safety testing. From competitive advantage to compliance burden. The strategy has done a complete 180, and nobody’s been asked to explain why.

The enforcement paradox

Let's say the bill passes. How exactly does the UK regulate frontier AI models?

Every major frontier model is built abroad. Anthropic, OpenAI, Google DeepMind’s parent, Meta — all American. DeepSeek is Chinese. The UK doesn’t control any of them. So what does “mandatory registration” actually mean?

Option one: you make foreign developers register and comply with UK testing requirements before offering their models here. If they refuse or it’s not worth the compliance cost, UK businesses lose access. In a market where AI adoption is becoming a competitive necessity, that’s not safety — it’s self-harm.

Option two: you regulate UK-based companies that deploy these models. But then you’re just adding cost to British businesses while the models themselves are built elsewhere with zero UK oversight. You’ve created friction without achieving safety.

Option three: you hope for international cooperation. The US under the current administration is actively deregulating AI. Good luck with that.

The bill targets “frontier models” — those exceeding certain computational thresholds. But those thresholds will be set by secondary legislation, which means they can be lowered without full parliamentary scrutiny. Today’s exemption is tomorrow’s compliance burden.

We've seen this film before

If this sounds familiar, it should. Europe already ran this experiment with data protection.

GDPR was going to set the global standard. It was going to protect citizens and create a competitive advantage for trustworthy European businesses.

What actually happened? Researchers at the National Bureau of Economic Research found that GDPR reduced the number of EU technology venture deals by 26% relative to the US. A separate analysis by the American Action Forum found that venture capital investment in small and micro companies decreased by $3.4 million per week following GDPR’s introduction. The EU now invests five times less in private tech R&D than the US.

Now, correlation isn’t causation. Europe’s tech gap has many causes — capital markets, language fragmentation, risk culture, scale. But GDPR demonstrably didn’t help. It added cost without creating competitiveness. And the compliance burden fell hardest on the startups and small companies that Europe needed most.

The UK was supposed to be the alternative. Post-Brexit, the pitch was clear: lighter regulation, faster decisions, a better place to build. The AI Registration and Oversight Bill makes the UK look more like the EU, not less.

The honest counterargument

I should be fair. The case for some regulation isn't stupid.

Voluntary commitments did fail. When companies are under competitive pressure to ship faster, voluntary safety testing is the first thing to go. The Bletchley pledges had no teeth, and some companies acted accordingly. There are genuine safety concerns with the most powerful models — the potential for misuse in bioweapons research, large-scale disinformation, or critical infrastructure attacks isn’t fiction. And the bill is narrowly targeted at frontier models, not every chatbot and recommendation engine. That’s more sensible than the EU AI Act, which tries to regulate everything.

But here’s the part that’s easy to miss. The UK already has substantial regulatory coverage for AI. The Data Protection Act covers AI processing of personal data. The Equality Act covers discriminatory AI. The Online Safety Act covers AI-generated harmful content. The FCA, MHRA, and CMA all regulate AI in their sectors. Even AI Minister Kanishka Narayan acknowledged in January 2026 that existing rules already apply to AI systems.

So the question isn’t whether AI should be regulated — it already is. The question is whether another layer of cross-cutting compliance achieves anything the existing framework doesn’t. And given the enforcement problem — that you can’t meaningfully regulate developers in California from Westminster — the answer looks a lot like no.

Policy by press release

Perhaps the most telling detail is that the bill text doesn't exist yet. Peter Kyle announced the policy direction. He described the broad strokes. But there's no published legislation to scrutinise, no impact assessment, no detailed analysis of costs and benefits.

This is policy by press release. And it comes from the same government that simultaneously promotes AI Growth Zones — dedicated areas with lighter regulation to attract AI investment. The contradiction is stark. You can’t be the place that welcomes AI innovation and the place that regulates it more heavily at the same time. Investors notice.

What would actually help

The UK has strong universities, a deep talent pool, a globally recognised AI Safety Institute, and — until recently — a regulatory environment that was deliberately lighter than the EU's. That's a genuine competitive advantage — and it's exactly the environment needed for building AI-first businesses. The government should be doubling down on it, not dismantling it.

Scale the AI Growth Zones already being rolled out. Invest in compute infrastructure. Use public procurement to drive AI adoption across government services. Strengthen the AI Safety Institute’s research and advisory role — but keep it as a centre of expertise, not a compliance department with enforcement powers that deter the very investment the UK needs.

The answer to AI safety concerns isn’t a registration form. It’s competence — understanding the technology well enough to know where genuine risks exist and where regulation is just friction dressed up as protection.

The real question

The UK is falling behind. Growth is stagnant. Investment is the lowest in the G7. Real wages haven't moved in nearly two decades.

AI is one of the few genuinely transformative technologies that could change that trajectory. Not in a decade — now. The companies, the talent, and the infrastructure exist. The question isn’t whether AI will transform the economy. It’s whether the UK will be building that future or watching it happen from behind a compliance form.

The opportunity here is speed. Move faster than the EU. Move faster than the bureaucracy. Make the UK the place where an AI company can go from idea to product without hiring a regulatory affairs team first. That’s how you turbocharge growth — not by adding another layer of oversight to technology you don’t control, built by companies you can’t regulate, in countries that aren’t playing by your rules.

Peter Kyle says the British public expects proper oversight of powerful AI systems. He’s probably right. But the British public also expects wages that grow, an economy that works, and a government that knows the difference between protecting people and protecting itself from headlines.

Right now, this bill looks a lot more like the latter.