AI regulation splits Trump administration as frontier model order stalls
Trump postponed signing an executive order on voluntary AI model review after saying he "didn't like certain aspects" of it, citing concerns about U.S. competitiveness with China.
Objective Facts
President Trump postponed plans on May 21, 2026, to sign an executive order establishing a voluntary review process for artificial intelligence models before they're released, saying he "didn't like certain aspects" of it. The order would have created a voluntary review process for frontier AI models prior to public release, with the NSA setting which systems counted as "covered frontier models". The intervention was led by former Trump AI czar David Sacks, who called the president directly the morning of the planned ceremony after participating in the EO review earlier in the week; Elon Musk and Mark Zuckerberg also pushed back. OpenAI's chief lobbyist Chris Lehane had been broadly supportive of "collaborating with the government on AI safety"; other industry executives had pressed for a 14-day rather than 90-day model-sharing window. China issued a 2026 legislative work plan outlining comprehensive AI legislation while Washington cancelled the AI oversight ceremony.
Left-Leaning Perspective
Analysis in Transformer News identified a faction within the Trump administration—likely including Treasury Secretary Scott Bessent and Chief of Staff Susie Wiles—that is "seriously grappling with frontier model risks" but "cannot convince Trump to listen to them — or his own voters — over the siren song of Silicon Valley money," with internal briefings showing officials' "barely disguised contempt for Sacks." Sacha Haworth, executive director of the technology watchdog Tech Oversight Project, said the White House "keeps handing Dems opportunities on silver platters to point out how Big Tech CEOs being in bed with lawmakers means the American people lose." New York Governor Kathy Hochul framed the state-level AI safety victories as defeats for "AI oligarchs" and Trump's attempt to "stop RAISE through executive action greenlighting a Wild West for AI." The analysis suggests the administration's capitulation to billionaire founders reveals the hollowness of its claims to represent working-class interests against corporate power. Left-leaning critics note the dramatic reversal from February 2025, when Vice President JD Vance warned at the Paris AI Action Summit that "the AI future is not going to be won by hand-wringing about safety" and suggested excessive regulations might "kill a transformative industry," making the sudden concern about Mythos capabilities appear nakedly self-interested rather than principled. Left coverage largely emphasizes that a voluntary testing framework is minimal and that the order's cancellation shows the administration cannot act against industry pressure even on national security grounds.
Right-Leaning Perspective
Dean Ball, one of the primary authors of the White House AI Action Plan, stated on social media that "the burden of frontier AI governance is going to rest principally on the private sector," signaling conservative approval of self-regulation. Daniel Castro, president of the Information Technology and Innovation Foundation, praised the postponement, writing that "hitting pause on the proposed AI executive order will give the administration more time to avoid unintended consequences," and cited the EU's "stringent AI regulatory measures" as a cautionary example of overregulation. Andrew Lokay, a senior research analyst at Beacon Policy Advisors, characterized Trump's decision as "a sign of the continuing influence of Silicon Valley with the Trump administration," though questioned whether Sacks now has excessive influence over AI policy. Right-leaning analysis emphasizes preventing regulatory overreach and maintaining U.S. competitive advantage, while some conservative figures worry that a voluntary framework could slide into de facto licensing. Sources familiar with Trump's thinking told Axios that the president "just hates regulation" and viewed the order as "something doomers wanted." Right coverage frames the cancellation as consistent with Trump's deregulatory agenda and highlights concerns about China competition rather than portraying it as a victory for Big Tech lobbying.
Deep Dive
The core story is a dramatic split within the Trump administration between accelerationists around David Sacks and Mark Zuckerberg who oppose any pre-release government review of frontier AI models, and a smaller faction led by Treasury Secretary Scott Bessent and Chief of Staff Susie Wiles concerned with national security implications of models like Anthropic's Mythos. Trump cancelled the May 21 signing after consulting with Sacks, Musk, and Zuckerberg overnight, saying he "just hates regulation" and wanted to avoid slowing U.S. AI development. The order divided the AI industry itself—OpenAI's Chris Lehane publicly supported collaboration on government AI safety testing, while Musk's xAI and Zuckerberg's Meta opposed it as an innovation blocker. The plan Trump killed was "not aggressive regulation" but "close to the gentlest version imaginable"—a purely voluntary 90-day review framework for frontier models before public release. The order was "not killed on substance over weeks of negotiation; it was killed in 12 hours by a single phone call." The cancellation means the United States currently has no formal pre-release oversight mechanism, despite the Mythos case having demonstrated that frontier model capabilities can reach security-relevant thresholds without any established government review process. What each side gets right: Trump administration accelerationists correctly note that premature or poorly-designed regulation could create competitive handicaps against Chinese AI development. OpenAI's support for the order reflects legitimate industry recognition that some voluntary, predictable oversight framework is preferable to reactive, chaotic regulation later. What they miss: The accelerationist argument ignores that Mythos demonstrated the novel ability to autonomously discover thousands of severe cyber vulnerabilities, prompting Vice President Vance to warn that "a bad actor could use Mythos to target various cybersecurity vulnerabilities." Anthropic's April 2026 announcement that it had developed a model "too dangerous for unrestricted release" was unprecedented—a frontier AI laboratory publicly acknowledged building a model it chose not to release. Safety advocates underestimate the genuine complexity of balancing security with innovation speed in the AI race. What happens next remains unclear: the federal government has "drafted, staged, and abandoned a frontier-model review process in the span of a single afternoon," suggesting "the regime is coming but the timing is hostage to politics," with "agencies, definitions of a covered model, and the 90-day idea" all "written down somewhere in a White House drawer" and likely to "come back, probably after the next model does something frightening in public."