Donald Trump Won't Slow AI Down

The president dismissed a rare joint plea from the CEOs of Anthropic, OpenAI and xAI to ease the pace of AI development. Uncomfortable as it sounds, he may be reading the stakes correctly.

Credit: Brendan Moran / Sportsfile

President Donald Trump on Sunday brushed off a joint warning from three of the most powerful people in artificial intelligence, telling reporters at his golf resort here that the United States has no business easing off the accelerator while China is still on the track.

"We're leading China in AI. We're the most sophisticated country in the world, and frankly I want to keep it that way, because whoever wins AI wins," Trump said, asked whether the industry should slow down or face tighter regulation.

According to a BBC report, he didn't rule out limits entirely:

"We could put guardrails. We can do this and that," he said. But he waved off the researchers and executives pressing the case for caution as "very negative forces" pushing scenarios that "won't happen."

The president's remarks were his first public response to a joint statement issued Saturday by Anthropic's Dario Amodei, OpenAI's Sam Altman and xAI's Elon Musk, three men who are otherwise locked in one of the fiercest corporate rivalries in American business.

Amodei laid out a three-step framework meant to pace the rollout of more capable models and buy safety researchers time to build in protections. Altmans signed on within hours, saying the pace of development had become "a primary topic of discussion" inside OpenAI. Musk backed it too.

The alignment was notable mostly because it doesn't happen. These are three companies burning through hundreds of billions of dollars to out-build one another, each betting its future on being first. For their chiefs to agree publicly on anything, let alone on hitting the brakes, is the kind of thing that normally moves markets and makes Washington sit up.

It barely dented Trump's Sunday.

What set it off?

The CEOs' statement followed a rough two weeks for the industry's safety argument. Jacob Coxon, a researcher who spent three years working on model pretraining at both Anthropic and OpenAI, quit Anthropic last Tuesday with a lengthy public post accusing both companies of "racing straight to self-improving superintelligence and gambling with our lives."

He said executives at both firms privately fear their own creations could kill people by the end of the decade, whatever they say for the cameras. His posts were read by tens of millions within a day.

Anthropic had already disclosed that its Claude models were used in attempts to help build bioweapons, spy on Ukraine and run dating-app scams, findings the Pentagon cited earlier this year in its own warning about the company's systems.

Senator Bernie Sanders said he'd introduce legislation to pause development of superintelligent systems altogether an he even sent a letter to top AI developers for slowing down the AGI development.

In the House, Democrat Josh Gottheimer and Republican Mike Lawler introduced a bipartisan bill aimed at keeping AI systems under human oversight.

Speaker Mike Johnson pushed back on moving too fast toward regulation, warning it could choke off American innovation. Democratic leader Hakeem Jeffries called for the opposite, urging faster federal action to protect the public.

None of this is happening in a vacuum ahead of November's midterms, which will decide control of Congress. Musks political operation, America PAC, has already spent millions backing Republican candidates this cycle, and AI money more broadly has flowed into a fight where Democrats are testing whether the White House's hands-off approach to AI regulation can be turned into a liability.

Inside the administration, the fault line has a face: David Sacks, the White House's AI adviser, has spent weeks sparring publicly with Amodei, accusing Anthropic of pushing state-level AI rules that would lock out smaller rivals while protecting its own position, a strategy he's called regulatory capture dressed up as safety.

Amodei has denied it, saying Anthropic wants the same outcome as the administration: American dominance in the field. Sacks has called the loudest AI warnings "overheated" and argued the technology that creates the risk is also the tool that solves it.

The case Trump isn't making well, but might be right about!

Set aside the golf-course delivery and the "negative forces" line, which does no favors to anyone actually worried about what's coming. The strategic logic underneath it is harder to dismiss than the phrasing suggests.

If artificial general intelligence arrives this decade, as Amodei himself has argued it might, the country that gets there first doesn't just gain a commercial edge. It gains a tool that can accelerate weapons design, code-breaking, propaganda and surveillance faster than any rival can respond to it.

A world where Beijing crosses that line first isn't a world with slightly worse AI safety standards. It's a world where an authoritarian government holds the kind of structural advantage that the United States has spent 80 years ensuring no adversary gets.

Washington didn't stop building nuclear weapons in the 1940s because a runaway arms race carried catastrophic risk. It built faster, precisely because the alternative, letting someone else get there first, was judged the greater danger and it proved it.

The AI race here is being argued in those same terms inside the administration, whether or not anyone there says so directly but some of the insiders are saying it is way larger than that.

That doesn't make the extinction-level warnings from people like Coxon irrelevant.

It's the argument critics of Trump's position make in return: a race run by two or more labs cutting corners under competitive pressure is not obviously safer than a single actor moving carefully, and an intelligence explosion doesn't ask permission from the country that triggers it.

Losing control of a powerful model doesn't stay inside national borders. On that point, the slow-down camp has a case Donald Trump's Sunday remarks didn't really answer.

What's harder to argue with is the asymmetry underneath both risks.

A US-led AI program that goes wrong is a catastrophe the world would face together, with American labs, American regulators and American courts still standing at the end of it, however damaged.

A Chinese-led AGI program that goes right, for Beijing, is a permanent shift in who sets the terms for everyone else.

Trump's bet, stripped of the rhetoric, is that the second outcome is the one Washington genuinely cannot survive being second in, even if it means tolerating the first risk to avoid it.

Anthropic, OpenAI and xAI have not said whether they'll act on their own framework without government backing.

As of Sunday night, the White House had not indicated it would seek new restrictions on frontier model development.