Something genuinely unusual happened this week and it got buried under the noise of the reaction.
An industry asked to be regulated, and the government said no.
On Saturday Dario Amodei published his pacing essay. By Sunday Sam Altman had agreed, writing that pacing does not mean stopping, that progress should be slower than it otherwise could be, and that no amount of American competitive pressure justifies recklessness or letting capabilities get ahead of alignment and monitoring. He said OpenAI would welcome a federal framework setting consistent safety requirements for frontier AI, and added that the company does not believe it needs to wait for an antitrust exemption or legislation to begin the work.
On Monday the President called the whole thing a hoax. AI taking over the world and destroying humanity, he wrote, is a HOAX, no different from the others America has been forced to endure. There is a SICK conspiracy going on against AI and data centres, and the only party happy about it is China. The only control or guardrails AI needs is a strong and smart president, and the USA has one in spades. Don’t kill the Golden Goose.
The Vice President was cooler and more interesting. JD Vance said the administration is concerned but wants to regulate smartly, then added that he feels a little weird about so many frontier AI companies coming to government and begging to be regulated. It feels, he said, a bit like a Trojan horse.
Set the AI question aside entirely for a moment, because the commercial pattern underneath this is much older and it is worth seeing clearly.
Firms almost never ask for rules they do not benefit from. Look at any mature industry and you will find safety standards, licensing regimes and codes of practice that the incumbents fought for, and the reason is rarely virtue and rarely pure cynicism either. It is usually both at once. A rule you can already comply with is a rule your smaller competitor has to build toward, and a cost that is trivial at your scale is fatal at theirs.
Coffee has the tidiest example in commerce. Food hygiene ratings, allergen labelling, traceability requirements, certification schemes. Every one of them protects the public, genuinely and measurably, and every one of them is easier to satisfy if you already have a compliance officer. The chain welcomes the inspection. The stall on the corner discovers that meeting the standard costs more than the standard’s designers ever spent.
So Vance’s Trojan horse remark is a real argument and not a cheap one. Technology analysts have made the same point this week, that safety warnings can invite regulation which then functions as a competitive moat. It deserves to sit in the room.
What it does not do is settle the question, because the argument cuts in a direction its users rarely follow. If you believe the incumbents are asking for rules that suit them, the response is to write rules that do not suit them, not to write no rules at all. Declining to regulate does not deny the incumbents an advantage. It hands them a different one, which is the absence of any obligation whatsoever while they hold the largest models, the most compute and the most capital.
Trump’s own framing gives the game away, and it is not really about safety. The AI buildout has accounted for a large share of economic growth this term, and data centre construction has carried the manufacturing figures. Golden Goose is an economic sentence.
There is a meeting with AI executives being scheduled this week, according to House Speaker Mike Johnson.
The people holding the valve are the ones asking somebody to take it.
