MW Here's how far the Trump administration's 'startling turn' on AI regulation might go
By Hannah Pedone
After promising a hands-off approach to AI innovation, the government may tighten oversight on tech giants' models
The Trump administration once decried "onerous" control of AI development. Now, the government may become more involved in vetting AI models.
In President Donald Trump's first month in office, he signed an executive order revoking a 2023 Biden administration regulatory process created to protect Americans and U.S. national security from risks posed by artificial intelligence.
At the time, the White House said that the Biden-era order hindered AI innovation and imposed "onerous and unnecessary government control over the development of AI."
The executive order was one of several signals during that period - including prominently featuring major tech CEOs at Trump's inauguration three days before the AI policy was announced - that the Trump administration was going to take more of a hands-off approach to tech giants and their products. But now, the White House is reportedly considering getting the government more involved, at least when it comes to AI.
The Trump administration is weighing taking a more aggressive approach to regulating AI, which could include an executive order to increase oversight over artificial-intelligence models before their public release, according to a Monday report by the New York Times, which cited U.S. officials and individuals briefed on the deliberations.
A White House official told MarketWatch that discussion of a potential executive order, which could include a formal government review process for AI models, is "speculation," and that any policy announcement will come from the president.
Still, the Trump administration is moving forward on a plan to have tighter oversight of AI models through voluntary agreements. On Tuesday, the Center for AI Standards and Innovation within the Department of Commerce announced agreements with Alphabet's $(GOOG)$ $(GOOGL)$ Google DeepMind, Microsoft $(MSFT)$ and xAI, whereby the center will conduct "pre-deployment evaluations" of the companies' models.
The reports and the deals with AI giants come as the public is growing increasingly skeptical of AI and as concerns over potential job losses and expanding data centers tied to the technology's development are becoming more politically salient. In addition, some AI innovation has the potential to pose national-security risks. But it remains unclear to what extent the government will be involved in monitoring the companies as they innovate at a rapid pace.
Back to the Biden administration approach?
Darrell West, a Senior Fellow at the Center for Technology Innovation at the Brookings Institution, a think tank, told MarketWatch that it appears Trump may be returning to the Biden administration's approach of AI oversight.
"It is a startling turn from his libertarian stance of the past year where he allowed tech companies to do what they wanted without hardly any government oversight," he said.
Trump's top AI adviser David Sacks, an entrepreneur and investor who advocated for a hands-off approach to AI regulation, announced in March that he is leaving the White House, a move West says will free the administration to move toward more regulation of AI models.
See also: How this Japanese toilet maker became an unlikely AI winner
James Lewis, a senior adviser at the Center for Strategic and International Studies, a national security-focused think tank, and a former diplomat, told MarketWatch that while the Trump administration has made a "major change in direction" with the CAISI partnership announcement, it's unlikely that this White House will go as far as the Biden administration in monitoring the industry.
The Biden White House was focused on a broader AI agenda that included safety and privacy as well as equity and diversity in AI models.
"This administration is a little allergic to government surveillance, so I doubt [an executive order] would go beyond national security," he said, explaining that any policy that would expand "intrusive government surveillance" would run into opposition within the administration.
Concerns over AI's ability to exploit vulnerable cybersecurity systems
Lewis said that the Trump administration's pivot was likely due to officials' alarm over Anthropic's release last month of Mythos, a model that can discover and exploit cybersecurity vulnerabilities.
Anthropic withheld the model's release to the public, but made it available to a select group of companies through Project Glasswing.
"I think a lot of it comes from Anthropic persuading [the Defense Department] that this was a risk that needed to be seriously considered," Lewis said, adding that the technology was the "biggest challenge to cybersecurity we've seen in a long time."
"It scared enough people in the administration that you were gonna have a cybersecurity disaster that they floated this idea."
That said, Lewis noted that the CAISI partnership is still "voluntary" and "non-regulatory" and that companies "can't be compelled to participate."
He expects the National Security Agency will be involved in the partnership, given the review process is geared toward national-security interests.
While details of the program are limited, Lewis said that a voluntary submission program, run by the Department of Commerce, could give companies who come forward to be vetted safe harbor from further prosecution and liability, through a measure akin to the Cybersecurity Information Sharing Act, whereby a company that comes forward with information about an incident has legal protections.
Still, industry proponents are skeptical of the move toward more government oversight in the space. While companies are unlikely to object to government access through the CAISI program, Daniel Castro, president of the Information Technology and Innovation Foundation, a pro-tech think tank, said he worries what steps the government could take after gaining access and how the companies may handle government demands.
He said that a sweeping pre-approval regime of AI models could hamper the industry and put the U.S. behind other countries in developing the technology.
"It effectively forces firms to seek permission to innovate, slowing development to the pace of Washington rather than the market," he said.
"Other countries will keep advancing, and open-weight models released abroad won't face similar constraints-undercutting the U.S. AI ecosystem."
Read more: Big Tech's $700 billion spending on AI this year is called the 'greatest capital misallocation in history'
-Hannah Pedone
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 06, 2026 16:35 ET (20:35 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments