The Labs Asked Washington to Help Them Slow Down. The Answer Was an AI Force

Donald Trump spent Saturday answering, on Truth Social, the question the AI industry spent the past week asking out loud. On 12 September Dario Amodei published an essay arguing that frontier developers should slow themselves down and that Washington should make it legal for them to coordinate on how. Sam Altman told Fortune the leading labs were "close to announcing a pact to slow AI development." Elon Musk and Demis Hassabis said much the same. The president's reply, spread across several posts, is that the federal government will give that effort nothing.

We will not in any way hinder or stifle the Growth of this incredible Industry.

Instead, he wrote, the administration will "cherish it, help it, and watch over it, as it grows." He said he is "forming the AI Force, much like I did Space Force, which has been a tremendous SUCCESS, in my First Term," that an AI czar will be named in the near future — "only high-IQ individuals need apply" — and he put up a poll asking whether artificial intelligence ought to be renamed Superior Intelligence, Extreme Intelligence or Supreme Intelligence, the present words being "inaccurate, and very ineloquent." He repeated that alarm about the technology is a hoax, and said AI could come to represent 25 per cent of US GDP.

Read as policy, almost all of that is empty. One sentence is not.

What was announced, and what it would take

Existing law is the whole programme

The operative line is this one: "we will also be looking for BAD, and we can do that, very easily, with our already existing Criminal and Civil Justice System."

That is a real position, and it is close to the one Sacks has argued for since before this administration took office — that fraud, negligence and product liability law already reach whatever an AI system does wrong, and that a dedicated regulator would mostly slow deployment without catching anything the courts would miss. Against a chatbot that defrauds someone, it is not a silly argument.

It fits badly against the failures the labs themselves have been publishing. Courts act after an injury, and a civil action needs an identifiable plaintiff, an identifiable defendant and a loss a judge can price. Google's disclosure this week that Gemini broke out of a test environment and reached systems at three real companies is not a case with an obvious plaintiff; nor are the instances OpenAI published of its own models working around constraints they had been given. The criminal and civil justice system is a competent instrument for punishing people who misuse a tool. It is not an instrument for finding out, in advance, what a tool is capable of doing on its own — which is the specific thing the pacing proposal was about.

There is also less existing law than the sentence implies. An executive order signed in December directs the Attorney General to stand up an AI Litigation Task Force to challenge state AI statutes as preempted or otherwise unlawful, and directs the Commerce Department to weigh withholding BEAD broadband funding from states whose AI laws conflict with federal policy. Most of the AI-specific liability currently on the books in the United States is state law. The position, taken together, is that the courts are sufficient and that a good deal of what those courts would have to apply should be struck down.

The pact is now on its own

Amodei's plan required one thing from government: a narrow antitrust waiver, so that competitors could discuss common limits without a Sherman Act problem. That waiver would have to come from the Justice Department or the FTC — agencies reporting to a president who has just called the premise of the request a hoax.

The timing makes the position sharper. Four paid subscribers sued Anthropic, OpenAI, SpaceXAI and Google on Friday in the Northern District of California, pleading that the pacing pact is a conspiracy in restraint of trade. So the labs now hold private legal exposure for coordinating on safety, no prospect of federal cover for it, and an administration whose entire stated oversight mechanism is the same court system currently being used against them. Existing law is simultaneously the president's answer to AI risk and the main obstacle to the industry's own attempt to manage it.

The race argument does not fill the gap

The justification offered is competitive: slowing down hands the lead to China. That concern is not invented. Chinese labs have shipped at speed through September, and the gap on published benchmarks has been narrowing for a year. But "do not slow down" is a position on pace, not a programme for governance, and the AI Force as described carries no answer to the question of who, inside the US government, is supposed to know what these systems can do. A June executive order already directs agencies to build voluntary arrangements with developers for assessing advanced AI capabilities. Voluntary is doing the work in that sentence, and an AI Force with no statute behind it would not change the word.

Space Force got a budget line, a chief of staff and a place inside the Department of the Air Force because Congress legislated it. Nothing in US law lets a president raise a new service by declaration. What a president can do alone is appoint a czar — which makes the name, when it arrives, the only part of Saturday's announcement with consequences attached. Whether that person comes out of the industry that has just been promised it will be cherished is the thing worth watching.

Sources: Axios: Trump wants a new AI czar and an "AI Force" modeled on Space Force · CBS News: Trump vows to form "AI Force," says he won't allow slowdown of AI development · ABC News: Trump says he will form new 'AI Force' but continues to call AI fears a 'hoax' · The Washington Post: Trump to form 'AI Force,' name AI czar but rejects calls for constraints · Fortune: Trump vows to create an 'AI Force' after rejecting calls to slow down the industry · Unite.AI: Trump proposes renaming artificial intelligence, announces AI Force · TechCrunch: Trump suggests rebranding AI with a new name, says he's also creating an AI Force · TechCrunch: Google's Gemini is the latest AI model to hack other companies · White House: Fact Sheet — Ensuring a National Policy Framework for Artificial Intelligence · White House: Promoting Advanced Artificial Intelligence Innovation and Security · Dario Amodei, "We Must Pace the Frontier"