AI executive action stalled by White House infighting

Disagreement among administration officials and a time crunch with President Trump's China summit are holding up efforts to launch a federal response to the next frontier of AI.

Why it matters: There's not yet any new federal AI regulation weeks after Mythos — Anthropic's most advanced model yet — threw Washington for a loop.

• Early talk of federal safety reviews of new AI models has slowed down as Trump administration comments have made it clear they aren't on the same page just yet.
• Leaders across the government are spooked by the capabilities of leading AI labs' latest models, but the realities of Washington are slowing down any immediate action.

Driving the news: AI is likely on the agenda for Trump's trip to Beijing this week, and the industry is anxiously awaiting new guidance in response to the release of the latest frontier AI models.

• One tech industry source told Axios the administration may wait on the outcome of the China summit before making any final decisions on AI.
• The global picture is also getting more complicated, with European governments seeking access to models like Mythos for their own security testing.

What they're saying: Administration officials had to walk back one of their own suggesting safety reviews like an "FDA for AI" — a sign of a lack of cohesion about what the government should do next.

• After National Economic Council Director Kevin Hassett said the White House was studying how "future AIs" may go through a process "just like an FDA drug," other officials were quick to tamp down his remarks, including White House chief of staff Susie Wiles and former AI and crypto czar David Sacks.
• "The real issue is not what the American labs do, it's the fact that Chinese models and other models that other actors could train are going to have advanced cyber capabilities within the next six months or so, and so we need to take steps now to harden our systems," Sacks, who currently co-chairs the President's Council of Advisors on Science and Technology, told Fox Business this week.
• Hassett later told CNBC that "nobody has an idea that we should do something like bring in a giant new bureaucracy to approve AIs."

Between the lines: "It's fair to say that there are some differing views in the administration about how to approach the issue of these highly capable AI systems," another tech industry source that works with the government closely told Axios.

• "What industry needs is some clarity going forward so that we know how to handle the next model breakthrough. This isn't a process we want to recreate on the fly every time there's new tech."
• A White House official told Axios: "The White House continues to balance advancing innovation and ensuring security in AI policymaking. Any policy announcement will come directly from the President. Discussion about potential executive orders is speculation."

The other side: "We don't have time to waste, to point fingers or to let bureaucracy get in the way of implementing a process here for for proper [AI] testing and communicating results of that of those tests, whether that's on the classified side or unclassified side," Rep. Josh Gottheimer (D-N.J.), who co-chairs the House Democratic Commission on Artificial Intelligence and the Innovation Economy, told Axios.

• "I think they can use executive action while [Congress] gets to a legislative solution, to put forth a process for testing these new models and making sure that the results are deployed," he said.

Behind the scenes: There's also a tug-of-war over where advanced AI testing should happen — in the more "civilian" space of the Department of Commerce or in the national security side of the government, another tech industry source told Axios.

Case in point: After a division of the Department of Commerce, the Center for AI Standards and Innovation, publicized new frontier AI testing deals, the website was taken down just days later.

• CAISI staff were told to take down the page with no explanation, a source familiar with the situation told Axios.
• The White House and Commerce declined to provide on-the-record comment on why the page was was pulled or whether the program is ongoing.

The bottom line: Industry sources are still expecting and hoping for new guidance on AI and cybersecurity, model testing and rules around how AI companies contract with the government.

• But momentum in Washington could run out quick.