White House AI Oversight Order Shelved as Senior Officials Split Over Regulatory Approach

President Donald Trump pulled back from signing a prepared executive order on artificial intelligence testing and oversight on May 21, halting a framework that had been under negotiation for months and exposing sharp disagreements among his senior advisers over the federal government's role in regulating frontier AI models. The White House postponed plans for Trump to sign the order, which would have established a voluntary review process for AI models before public release, after Trump said he had delayed the signing because he "didn't like certain aspects" of the order. The signing had been postponed several times prior to this latest deferral. Senior White House officials told reporters the order is not canceled, only delayed, though no rescheduled date has been set [POLITICO].

The shelved order would have drawn its authority from the President's broad Article II executive power over federal procurement and national security, building on the administration's January 2025 Executive Order 14179, which directed agencies to remove barriers to American AI leadership. One draft of the order was divided into two sections: one addressing cybersecurity and a second governing "covered frontier models," which would define the class of AI systems subject to a voluntary framework allowing up to 90 days of pre-release government review. The Department of Commerce's National Institute of Standards and Technology had announced earlier this month that major technology companies would share unreleased AI models with the government for national security and public safety evaluation, a voluntary arrangement that predated the formal executive order and suggested the administration was already constructing an informal review architecture around it.

Three distinct factions have emerged inside the administration on the question of how aggressively to regulate advanced AI. Trump abruptly scrapped the order on May 21 after former AI czar David Sacks intervened directly with the president. Sacks and his allies favor a less restrictive approach, arguing that excessive regulation could weaken American companies in the competition against China. Sacks, a Silicon Valley venture capitalist and the first person to hold the White House AI and crypto czar role, has said he views his position as a bridge between the technology industry and Washington and has previously celebrated the administration's week-one action rescinding the Biden administration's AI executive order. The president reportedly reversed course after speaking with industry leaders including Sacks, as well as SpaceX CEO Elon Musk and Meta CEO Mark Zuckerberg.

The second faction centers on the Pentagon. Defense Secretary Pete Hegseth and Undersecretary of Defense for Research and Engineering Emil Michael have pushed for tougher barriers around advanced AI models, citing fears that cutting-edge capabilities could be exploited by rivals or bad actors. The catalyst for that concern is Anthropic's Mythos model. Mythos Preview has shown groundbreaking ability to spot cyber vulnerabilities, and Michael has described the Mythos issue as "a separate national security moment where we have to make sure that our networks are hardened up, because that model has capabilities that are particular to finding cyber vulnerabilities and patching them." The Pentagon's posture toward Anthropic itself is more adversarial: the Department of Defense declared Anthropic a supply chain risk, meaning its technology purportedly threatens U.S. national security, after the two sides failed to agree on how Anthropic's models could be used by the agency. Anthropic sued the Trump administration in March to try to reverse the Pentagon's blacklisting, a case that remains active and creates a parallel legal proceeding running alongside the executive order deliberations.

A third group, described as closer to the middle, includes White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent, who have reportedly advocated for a regulatory framework closer to what the shelved executive order would have imposed [POLITICO]. Their position reflects the argument that some structured federal oversight, rather than either a hands-off approach or outright prohibition, best balances competitive and security interests. The White House has urged Congress to create a federal framework to preempt state laws, but procedural hurdles and party disagreements have blocked any legislation from clearing Congress. That legislative stalemate adds institutional pressure on the executive branch to act unilaterally, making the fate of the executive order consequential beyond the immediate White House dispute.

The breakdown sits within a broader deregulatory posture the administration has maintained since its first days in office. Pursuant to Executive Order 14179 of January 23, 2025, Trump revoked his predecessor's AI executive order and directed the administration to remove barriers to U.S. AI leadership. An AI Action Plan released last summer dismantled many regulations concerning AI research, reversing Biden-era norms that promoted a whole-of-government approach with federal involvement in AI governance. The Trump administration had taken a more hands-off approach to AI regulation until recently, when Anthropic unveiled Mythos, which it says can exploit cybersecurity vulnerabilities at an unprecedented pace, forcing the question of pre-release government review back onto the policy agenda. Whether the administration moves to sign a revised order, pursue voluntary industry agreements through NIST, or defer the question to Congress will determine whether the current impasse produces a durable federal framework or simply extends the regulatory vacuum that has governed frontier AI development since January 2025.