Washington weighs equity in OpenAI and Anthropic while rejecting rules — that is a shareholder, not a referee

President Donald Trump told TIME he will not nationalise America's frontier AI labs — but he might buy into them. Asked by Eric Cortellessa and Sam Jacobs why Washington had not taken shares in OpenAI and Anthropic the way it did with Intel, Trump answered: "I might. Maybe I could do that." He put the value of the government's 10% Intel stake at about $60 billion and pointed to the cut Washington takes from some Nvidia chip sales. TIME published the full transcript of the 28 September White House interview, along with a fact-check.

The context sharpens the question. In July, per reporting cited by TIME, OpenAI offered the US government a 5% stake; Anthropic and the administration have denied discussing any stake in Anthropic. So of the two companies named, one has reportedly already floated the idea and the other says the conversation never happened. The dinner Trump had with Anthropic chief executive Dario Amodei and his wife the night before the interview supplies the mood music: Trump said Amodei's views on AI safety are "much different, I think, than what is portrayed in the media." Asked about OpenAI's Sam Altman, he said he liked him and that he was smart — but did not say whether he trusted him.

In the same conversation Trump repeated his opposition to AI regulation, arguing it could put companies out of business. Set those two positions side by side and you get the defining oddity of the moment: the United States is weighing whether to become a shareholder in the frontier labs while explicitly declining to be their referee. Historically the bargain runs the other way. Public capital arrives attached to public obligations — disclosure, audit, safety conditions. Equity without rules inverts that. It gives the state a financial interest in the valuation of the companies it would otherwise have to police.

The enforcement tool Trump did endorse is discretionary rather than structural. Asked about reports that OpenAI's AI agents had breached federal government websites, he said companies are not allowed to do that and "could have penalties that are not going to be acceptable to them," adding that he did not have to destroy the industry. Note what that is: deterrence by presidential displeasure, case by case, with no published standard an engineer could design against. It also lands squarely on the risk we think is the real near-term one — autonomous agents acting where they shouldn't. That problem gets solved by isolation, logging and liability rails, not by the threat of unspecified penalties after the fact.

On infrastructure, the politics are already visible in the numbers. TIME cited polling showing 71% of Americans oppose building data centres; Trump countered that they lower local taxes, pointing to Loudoun County, Virginia, which he said made $1.5 billion from them this year. Both things are true and that is the problem: the fiscal windfall is hyper-local while the grid load, water draw and construction disruption are diffuse. One county's $1.5 billion does not answer a national supermajority. He also said he now calls the technology SI, for super intelligence, rather than AI, and that Xi Jinping agreed — and that he will rename the AI Force he announced, describing it as a group of experts who will advise him.

**Our reading.** A sovereign stake in frontier AI is not a bad idea on its face. If this technology produces the abundance its builders promise, broad public ownership of the upside is one of the few credible answers to the concentration of returns we keep warning about. The problem is sequencing. Take the equity first and refuse the rules, and you don't get a sovereign wealth fund — you get a state that is financially invested in not looking too closely. Add a rebranding exercise and an advisory body named at the president's discretion, and the governance architecture starts to look personal rather than institutional. The long arc still points where we think it points: models that compress drug discovery, extend healthy life, and eventually make scarcity negotiable. Nothing in this interview changes that trajectory. What it changes is who captures it. The rules being drafted — or pointedly not drafted — in the next eighteen months will decide whether the gains are distributed or simply owned, and "I might take a stake" is a far weaker commitment to the public than it sounds.