MW Trump's AI executive order was set to feature voluntary participation by companies. He has delayed it.

By Victor Reklaitis

The president says the delay is because he doesn't want to hinder American AI companies that are ahead of their Chinese rivals

The Trump administration's position on artificial intelligence has shifted in part due to controversies around Anthropic's Claude models.

President Donald Trump on Thursday had been expected to sign an executive order on artificial intelligence that would set up a framework for AI companies to voluntarily inform the federal government about new models before their release.

That approach looked likely to disappoint critics of AI, including Trump supporters who have lobbied for a tougher stance due to national-security concerns.

But then Trump told reporters at the White House on Thursday that he had postponed a signing ceremony for the order because he "didn't like certain aspects of it." He suggested that was not because the draft order was too easy on U.S. AI companies, but rather that it could hinder them.

The president said the U.S. is leading China in AI and he doesn't "want to do anything that's going to get in the way of that lead."

"We have a very substantial standard on AI, it's causing - it's causing tremendous good, and it's also bringing in a lot of jobs, tremendous numbers of jobs," he added. "We have more people working right now than we've ever had. I really thought that could have been a blocker."

Before Trump announced the delay for the signing ceremony, analysts had been focusing on whether the proposed order could deliver on its promises.

The effectiveness of the AI framework would depend in part on whether having voluntary participation leads to "meaningful advance access" for the U.S. government, Raymond James analysts led by Ed Mills said in a note.

A voluntary framework could work, given that AI companies want to be able to get government contracts and participate in other government partnerships, the analysts added. "We will be watching if participation becomes quasi-voluntary for labs seeking federal partnerships, procurement access, or deeper integration into government AI initiatives," they wrote.

The proposed framework had looked like it would disappoint some Trump supporters who have urged him not to make participation voluntary. A letter to Trump signed by former adviser Steve Bannon and Amy Kremer, chair of a conservative organization called Women for America First, called for policies "that require mandatory testing, evaluation, vetting and government approval of potentially dangerous frontier AI systems before they are deployed."

The Trump administration appeared to shift its position on AI in recent weeks, after pushing for a hands-off approach shortly after Inauguration Day last year. The Raymond James analysts said the administration's initial stance on AI was "move fast, trust the companies, and avoid constraints that could slow the U.S. against China," but "that posture began to fracture in the first quarter of 2026 during the Anthropic-Pentagon dispute" - and then an even larger catalyst was Anthropic's limited April release of Claude Mythos.

Anthropic, a top AI company known for its Claude models, had objected in the first quarter to having its products used by the federal government to develop autonomous weapons or for mass surveillance. Then in April, the company's Claude Mythos model was withheld from the general public because of its ability to autonomously find and exploit software vulnerabilities.

Read more: Here's how far the Trump administration's 'startling turn' on AI regulation might go

One expert said it's possible that the delay in the rollout of the order could stem from the administration not consulting more extensively with AI companies.

The review of the draft executive order "hasn't been a really transparent process," said James Lewis, a senior adviser at the Center for Strategic and International Studies, a think tank. Lewis said his understanding is that not all AI providers were consulted.

"When you assemble 15 CEOs and give them basically 12 hours' notice to sign off on a major EO, there's always a risk that somebody will say, 'I can't make it, or I won't do this, or you're making a mistake,'" Lewis told MarketWatch. He said that the Biden administration "went completely the other direction in terms of caution" when consulting companies on executive orders. "They showed it to everybody in D.C. at least a month beforehand, and they aired every concern, every question."

Stocks tied to the AI industry have been a major driver of the U.S. stock market's recent gains. One exchange-traded fund holding such stocks, the Global X Artificial Intelligence & Technology ETF AIQ, is up about 22% so far this year and has gained 54% over the past 12 months.

Hannah Pedone and Robert Schroeder contributed.

-Victor Reklaitis

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

May 21, 2026 14:04 ET (18:04 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.