US President Donald Trump's recent turn to a more regulatory approach with artificial intelligence is in stark contrast to what he once promised, and opposition is growing.
During his campaign for a second term in the White House, he said he would embrace AI innovation and put less regulation on private technology companies in the US.
"We will repeal Joe Biden's dangerous executive order that hinders AI innovation and imposes radical left-wing ideas on the development of this technology," read the 2024 Republican National Convention platform, largely written by Trump campaign staff members.
"In its place, Republicans support AI development rooted in free speech and human flourishing."
Now, those promises seem to be vanishing as AI language models become ever more powerful, so much so that even cyber-security experts are worried that simple prompts could expose vulnerabilities in the most vaunted computer systems
Several weeks ago, the US Department of Commerce forced Anthropic to disable access to its latest AI models, Fable 5 and Mythos 5, for "any foreign national, whether inside the US or outside".
Anthropic, which had claimed those models were finely tuned to reduce the risk of misuse, said the Commerce Department had mentioned a vague risk that the models could be compromised and fall into the wrong hands.
In a letter to Commerce Secretary Howard Lutnick and Cyber Director Sean Cairncross, at least 100 AI executives, cyber-security experts and academics expressed worries about the nature of the decision.
Anthropic ultimately complied with White House orders, although recent reports suggest some progress has been made in allowing limited distribution of Fable and Mythos.
On Friday, it became clear that OpenAI, a rival of Anthropic, was placed under similar pressure from the White House.
In a lengthy statement, the technology company said it would be introducing its latest AI models in a limited preview “for a small group of trusted partners”, at the request of the White House and out of an abundance of caution.
OpenAI, led by Sam Altman, added that it does not want this sort of introduction to become a long-term solution to the Trump administration's worries about powerful AI models falling into the wrong hands.
“We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks, while we work with the administration to develop the cyber executive order framework and a repeatable process for future model releases,” it said.
Opposition to the White House's abrupt policy shift has been swift. Consumers have been concerned about the White House ultimately deciding which citizens and companies can have access to AI models.
John Coleman, legislative counsel for Fire, the Foundation for Individual Rights and Expression, lamented the lack of disclosure from the Trump administration as to who could use them.
"The problem here is that we just don't know enough about what their decision-making process is," he said. "What factors are they using?"
Mr Coleman said Fire believes AI fits perfectly into the First Amendment category for freedom of expression.
"People who are developing it and using it retain their First Amendment rights," he said. "Those rights include the right to share and receive information, the right to speak anonymously, the right to be free from compelled speech, among others."
Mr Coleman also said that Fire is disappointed in the White House's AI turn from a constitutional perspective.
"We've been praised by both sides," he said, referring to Democratic and Republican parties. "We've also been scolded both by both sides. It comes with the territory."
David Sacks, who was Mr Trump's senior AI and cryptocurrency adviser, also appeared to disagree with the White House's recent approach.
"A year ago, President Trump declared that America was in a global AI race and that the way to win it was to be pro-innovation, pro-infrastructure, pro-energy and pro-export," he posted on X while sharing an article about China catching up to the US in AI.
"President Trump was exactly right. We deviate from that strategy at our peril."
Mr Sacks was not alone on the conservative spectrum in reacting to the White House's about-face on AI regulation.
"This latest US government action signals a change," Neil Chilson, former chief technologist for the Federal Trade Commission during Mr Trump's first term, and Adam Thierer, a senior fellow technology fellow at R Street Institute, wrote in on Substack in June about the White House's restrictions on Anthropic.
"It significantly escalates the centralisation of control over advanced computation in our country."
Mr Chilson and Mr Thierer had been very supportive of the Trump administration's less regulatory approach to AI.
Technology expert Dean Ball, who earlier in the year helped to craft Mr Trump's AI Action Plan and defended the administration, said in a recent social media post: "In a matter of weeks US federal AI policy has gone from implausibly libertarian to increasingly draconian and opaque."
