Trump’s big-government, interventionist AI policy is the one thing he’s done right

Chris Gay is a contributing columnist for The Globe and Mail. He is a former Wall Street Journal staffer and writes the newsletter Figure at Center.

For all his faults, let’s give Donald Trump credit for one thing: He has pulled America’s Republican Party a considerable distance from its much-cherished if largely hollow small-government “principle.” There may be no clearer example than his – and his party’s – eagerness to place government guardrails around artificial intelligence.

Not a moment too soon. Unreleased AI models have shown a disturbing inclination to escape experimental laboratories and hack into other systems independently of their creators. In April, Anthropic’s Claude Mythos escaped its “sandbox” – an isolated testing environment supposedly unconnected to the Internet – and e-mailed a company researcher.

Opinion: Locked out of Anthropic’s Mythos, Canada must fight for AI access in USMCA talks

Last month, two OpenAI models, one recently released and the other still under evaluation, hacked into the systems of Hugging Face, an AI-model hosting platform, and then a customer of cloud provider Modal Labs. More incidents followed.

The U.S. President signed an executive order on June 2, asking technology companies to voluntarily give the government a look at AI models before releasing them to the public. That action reversed the hands-off approach Mr. Trump backed just weeks earlier, when he cancelled the signing of an order that would have allowed the government to review new models. (On Tuesday, the administration exempted “open-weight” models from the June 2 order, those anyone can download.)

The first order, he said at the time, “gets in the way of – you know, we’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead.”

A little ideological perspective is in order here. Isn’t it supposed to be those meddling, namby-pamby liberals who want to regulate everything, pick winners and keep a Big Government jackboot on the private sector’s neck? What happened to the scripted banalities such as “regulation will impede efficiency, cost jobs and put the economy on the slippery slope to central planning” – the sort of thing you hear when government wants to constrain, for example, systemically risky financial institutions?

Of course, lots of folks who cosplay as small-government conservatives don’t really mean it; they simply know what sort of language the donor base demands. Much free-market evangelism amounts to cheap rhetorical weaponry aimed at vilifying government and titillating the sort of people who think Ayn Rand is the last word in economic philosophy. A favourite chestnut: anti-tax zealot Grover Norquist’s demand that government be rendered small enough “to drown it in a bathtub.”

Top artificial intelligence companies to meet with Trump after rogue AI agent hacks

But it’s easy to cite government interventions that self-anointed libertarians would be up in arms about trying to abolish – laws that keep pornography out of the public square, illegal drugs off street corners, or any public good. (Next time you meet one of these folks, ask if they’d like to liberate the crack cocaine and child pornography markets from the oppressive hand of Big Government.)

Untethered to anything resembling principle, Mr. Trump has no qualms about indulging what Republicans used to condemn as Stalinist industrial policy. You may recall this jewel from 2019: “Our great American companies are hereby ordered to immediately start looking for an alternative to China, including bringing your companies HOME and making your products in the USA.”

Read that again: “hereby ordered.”

Since January, 2025, the U.S. government has acquired stakes worth $27.6-billion in at least 37 companies, among them Intel Corp., Westinghouse Electric Co. and Nippon Steel Corp. As the libertarian Cato Institute rightly noted, congressional Republicans “would be foaming at the mouth were this occurring under a Democratic administration.”

No more. Republicans generally favour AI regulation. The late South Carolina Sen. Lindsey Graham, a prominent Republican and A-team Trump toady, wanted a national standard for regulating AI, as opposed to the state-level regulations now emerging. He was right: Dividing authority over a borderless technology among 50 states would mean chaos and regulatory arbitrage. (Canada, of course, has a similar challenge.)

Mr. Trump is a creature of mood and impulse, and he may have nefarious ulterior motives for the June 2 order. But it is actually prudent. AI does pose unprecedented and unknowable risks – broadly known as the “alignment problem” – that it would be reckless to leave to self-interested private actors.

The order quite reasonably noted potential AI threats to national security and to children. It wants Congress to establish sandboxes, make federal data sets accessible to industry and academia, and “support development and deployment of sector-specific AI applications through existing regulatory bodies.” All laudable objectives, but note one thing: They are not the pursuits of the free-market, but of the state.

Fundamentalists who condemn government but fear AI might want to reconsider what they wish for. They might also stop to appreciate the luxury of knowing they’ll never have to live in the world they propose.