The European Union and the United Kingdom have spent years carefully building out frameworks for testing and regulating powerful AI models. The United States, meanwhile, elected a guy who ripped up the existing playbook on his first day back in office and is now, apparently, improvising. A deadline for the Trump administration to set actual rules is looming, and the architects of the European approach have some thoughts.
What Trump Actually Did
When Donald Trump returned to the White House for his second term, one of his first moves was to scrap the Biden administration's AI strategy entirely. Gone. Done. The prior framework, whatever its flaws, at least existed. Trump replaced it with what his team cheerfully described as an anti-regulatory regime, which is a polished way of saying they decided the answer to "how do we handle the most consequential technology in human history" was "fewer rules."
According to Axios, this was all happening while AI models were becoming increasingly powerful at a pace that was outrunning everyone's assumptions. The models got stronger faster than expected. The regulatory void got bigger faster than expected. And now there is a deadline on the calendar for the administration to actually produce a framework, which means somewhere in Washington right now, people are furiously Googling what the Europeans did.
What Europe Actually Built
Here is the thing about the EU and UK approaches: they did not materialize overnight. Axios reports that U.S. allies have been grappling with the exact same AI safety questions the Trump administration now faces, for years. The difference is they kept working on answers instead of declaring the question itself to be regulatory overreach.
The EU's AI Act, the most comprehensive regulatory framework attempted anywhere in the world, took years of negotiation, revision, and political combat across multiple member states with competing interests. The UK, post-Brexit, went its own direction with a more flexible, principle-based approach centered on sector-specific regulators. Neither system is perfect. Both systems are real. That distinction matters more now than it did six months ago.
The architects of those European systems, according to Axios, are watching the American deadline with a mix of interest and something that probably looks a lot like exhaustion. They have been down this road. They know how hard it is to get this right even when you are trying.
The Gap Between "Anti-Regulatory" and "No Rules At All"
There is a version of the Trump approach that makes a coherent argument. You could, in good faith, contend that heavy-handed preemptive regulation of AI could kneecap American innovation and hand competitive advantage to China. That is a real concern that serious people hold. The problem is that "anti-regulatory" and "no coherent framework whatsoever" are not the same thing, and the Trump administration has been treating them as interchangeable.
As AI systems get more powerful, the pressure to have some set of rules becomes harder to wave away. Not because regulators are naturally right about everything, but because powerful autonomous systems operating in financial markets, healthcare, critical infrastructure, and military contexts without any defined boundaries is not a freedom agenda. It is a liability waiting to introduce itself.
Axios notes that even within the administration's own framing, a deadline now exists. Which means someone, somewhere, had to admit that the vacuum was not sustainable. That is progress, technically. Extremely minimal, arrived-at-the-last-possible-moment progress.
Why the American Decision Actually Matters for Everyone Else
The EU and UK did not build their AI frameworks in isolation from the United States. American companies, American capital, and American-developed models are woven through the global AI ecosystem so thoroughly that whatever Washington decides ripples outward immediately. The Europeans know this, which is part of why Axios reports they are paying close attention to wherever the U.S. lands.
But here is what the European architects are also saying, according to Axios: whatever the U.S. produces as a framework would only be a start. Not a finish line. Not a solved problem. A starting point that will require constant revision as the technology keeps moving. The gap between where current AI capabilities are and where they will be in eighteen months is the whole problem. Any static framework becomes obsolete almost immediately. What matters is whether you have a functioning system for updating it.
The United States, having spent the first stretch of this administration dismantling the previous system rather than improving it, is not starting from a strong position on that front.
The Clock Is Actually Ticking
Deadlines have a way of concentrating minds, and the Trump administration now has one. What exactly emerges from that process is genuinely unknown. The optimistic reading is that the urgency produces something workable, something that acknowledges safety without strangling development, something the rest of the world can at least coordinate around.
The less optimistic reading is that the administration produces a framework shaped primarily by the preferences of the companies that spent heavily to elect it, wraps it in language about American AI dominance, and calls it a day. The European experience suggests that industry input in AI rulemaking is valuable and necessary. It also suggests that industry input as the only meaningful voice in the room tends to produce frameworks with interesting gaps exactly where the risks are highest.
Axios makes clear that U.S. allies are watching. The question is not whether America sets rules. At this point, some kind of framework is coming. The question is whether those rules are actually designed to address the hard problems, or whether they are designed to look like they do.
The Dingo Take
Scrapping your predecessor's AI safety strategy on day one is not a policy. It is a gesture. A vibe. A way of communicating tribal identity through executive action without doing any of the hard thinking that actual governance requires. The Biden framework had real critics and real weaknesses. The correct response to a flawed policy is to improve it, not to delete it and then spend a year and a half with nothing before a deadline forces your hand.
The Europeans, for all the grief they catch about bureaucratic overreach, at least took the question seriously enough to engage with it. The UK's approach is genuinely different from the EU's, and that difference reflects legitimate disagreement about how to balance safety with flexibility. That is what policy debate is supposed to look like. What it is not supposed to look like is one of the most powerful governments on earth treating "we ripped up the old rules" as a sufficient answer to questions about autonomous systems making consequential decisions at scale.
Here is the uncomfortable truth sitting underneath all of this: AI is not going to wait for the Trump administration to finish deciding how it feels about regulation. The models are getting stronger right now, today, while the deadline clock runs. Europe spent years building its frameworks and still says what the U.S. produces would only be a start. The United States is showing up to that starting line having spent the first leg of the race standing still. Whatever comes out of this deadline, the rest of the world is already ahead.
Comments