Here's the unpopular take that nobody in Washington seems willing to entertain: restraint, not speed, may be the smarter strategy here.

Watch the regulatory response to artificial intelligence over the past eighteen months and you'll see something that should concern even those who desperately want oversight. Policymakers are moving with the urgency of people who haven't yet figured out what they're regulating. Lawmakers are drafting "kill switch" bills. Treasury analysts are flagging systemic risk. Lawmakers are repricing frontier AI development. The pressure is on to do something, anything, immediately.

The problem is that speed in regulation usually leaves wreckage.

We've seen this movie before. Financial regulators moved fast after 2008 and locked in rules based on incomplete understanding of what went wrong. Environmental rules drafted in the 1970s created perverse incentives that took decades to untangle. Tech regulation in the EU happened so quickly that nobody has yet agreed on whether GDPR actually worked or just created theater.

The difference this time is that we're regulating something we're still actively learning about. The capabilities of large language models, their actual failure modes, their real economic impact, the ways they'll integrate into infrastructure—all of this is still being discovered in real time. We're writing permanent rules based on what amounts to preliminary evidence.

Consider what's happening with data centers. Companies want to build more of them. Some lawmakers want to restrict where they can go. This seems reasonable on the surface until you realize that nobody has perfect foresight about where AI will be most beneficial to locate, what the actual energy demands will look like in three years, or whether this approach will simply push development offshore. We're making spatial decisions based on anxiety rather than analysis.

The "kill switch" framing is even more revealing. It sounds serious and responsible. It's actually the regulatory equivalent of a panic button. A policy built on the assumption that AI could go catastrophically wrong at any moment, requiring instant shutdown, is a policy built on speculation, not evidence. It also creates perverse incentives. If companies believe a kill switch is coming, they have reason to hide problems rather than surface them. They have reason to avoid transparency.

Some will say that the pace of AI development justifies rapid regulation. Fair point. But faster regulation doesn't automatically solve faster development. It often just means slower innovation across the board, including the innovations we want. It means capital flowing to jurisdictions with lighter regulatory hands. It means brain drain and competitive disadvantage.

None of this is an argument against regulation itself. AI will need rules. The question is whether those rules get built carefully or hastily.

What would restraint actually look like? It would mean establishing strong reporting requirements before mandates. It would mean running regulatory pilots that last long enough to yield data. It would mean letting some states and countries experiment with different approaches rather than locking in federal rules that might be wrong. It would mean regulators actually understanding what they're regulating before they regulate it.

This approach is unfashionable because it requires politicians to sit with uncertainty for a while. It looks weak compared to the symbolic force of a "kill switch." But the actual weakness is writing rules you'll have to rewrite in five years because the technology rendered them obsolete or counterproductive.

The smartest move isn't always the fastest one. Sometimes it's the one that leaves room for course correction.