TL;DR: Et tu, Donald? According to a New York Times report, the Trump administration—which came into office promising major AI deregulation—is now backtracking. The White House is eyeing everything from government reviews to safety testing and working committees, moves that raise thorny questions about national security and global competitiveness and may irk the accelerationist-at-all-costs tech leaders who backed Trump’s second term. What’s in the works: Likely via executive order, the administration is reportedly considering creating a working group of tech leaders and officials to help design a formal AI vetting process. White House officials briefed Anthropic, Google, and OpenAI execs last week on early plans, according to the NYT, though one government official called the EO talk just “speculation” and another source told Axios, “There are still real tensions being worked through internally.” Central to the plans? Establishing a formal government review and safety testing of AI models before they’re released, with the Pentagon reportedly tapped to lead that testing. (The UK government is working on something similar.) Meanwhile, major US AI labs, including Google, Microsoft, and xAI, have already voluntarily agreed to give the government advance access to unreleased models, per Bloomberg. Saying the quiet part out loud: Anthropic’s much-discussed Mythos model (which could cause a cybersecurity “reckoning”) and OpenAI’s GPT-5.5 model likely forced the White House’s 180. Also worth noting: Much of what’s now being floated—safety evaluations, reporting on risky models—was originally in the Biden-era guardrails that Trump unwound last year. Internally, there’s also been a changing of the AI guard: Former AI czar David Sacks, who championed deregulation, left in late March; White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent have since filled his role and are seen as more hands-on and security-minded than Sacks. Pushback incoming: Oh, to be a fly on the wall at Marc Andreessen’s office right now. Per the NYT, many tech execs are predictably frustrated at the prospect of more regulation, arguing oversight could slow US progress in its AI innovation race against China—which was Trump’s original rationale for deregulating AI. China, meanwhile, already requires AI models to be approved by regulators before launch. Bottom line: The administration that came in promising to let AI rip is now asking how to rein it in—but whether oversight will look like a real check or a courtesy preview is still an open question. —WK | 
