Three Tech Executives Called Trump. Hours Later He Walked Away From His AI Order.
Trump postponed signing a landmark AI executive order May 21, hours before the ceremony was set to begin. He had already heard from three tech executives who wanted it stopped.
Elon Musk, Mark Zuckerberg, and David Sacks each called the White House between Wednesday night and Thursday morning, persuading the administration to pull the order from the signing ceremony, according to Semafor. Some executives had already flown to Washington when the postponement announcement came through, LiveMint reported.
The order would have created a voluntary framework under which AI labs share their most advanced models with the government at least 90 days before releasing them publicly, Axios reported. The point was to give U.S. agencies a window to assess national security risks from frontier AI systems before those systems reached the public. Industry groups pushed to shorten that window to 14 days, according to Politico. The White House text ultimately dropped the pre-release review requirement entirely, stating that nothing in the order authorizes mandatory governmental licensing or preclearance, LiveMint wrote.
Trump said afterward he postponed the order because he did not like certain aspects of it and because it could be a blocker to the U.S. leading China in AI, NBC News reported.
The three executives who called the White House had competing interests in the outcome. OpenAI, which has backed federal AI regulation, supported the order's general structure, Politico reported. Musk and Zuckerberg opposed it. Sacks, the former AI and crypto czar who helped draft earlier administration AI policy, told Trump the voluntary framework might eventually become mandatory in practice, raising concerns about future regulatory creep, Politico reported. He placed that call to Trump without telling his own staff first.
The contradiction at the center of this episode is that the industry used its access to eliminate an oversight mechanism before it ever took effect. The voluntary pre-release review was not mandatory. It carried no licensing requirement, no hold period, no enforcement mechanism. But it was a structure: a way for frontier AI labs to build safety thinking into the release cycle before models reached billions of users. The three executives with the most direct line to the president eliminated it in a matter of hours.
What this episode decided, beyond the fate of one order, is the question of who governs frontier AI development in the United States. The answer the White House delivered, with industry input, is that the labs govern themselves. The executives who called to complain were not proposing an alternative framework. They were making the case that no framework was needed.
That answer puts the U.S. in a unusual position globally. China is moving toward stricter oversight. Beijing issued rules in April requiring AI companies to establish internal ethics review committees, Artificial Intelligence News reported. The State Council has listed comprehensive AI legislation for the third consecutive year at the National People's Congress, SCMP reported. The United Kingdom has been stood up its AI Safety Institute with government-backed evaluation capability. Canada has proposed the Artificial Intelligence and Data Act, which would create mandatory impact assessments for high-risk AI systems. The EU's AI Act is already in force, with compliance deadlines rolling out through 2027. These efforts are imperfect and contested. But they are efforts.
The U.S. just passed on building any equivalent into federal practice. Aviation regulated itself through the 1950s and into the 1970s, until a series of accidents forced the creation of the Federal Aviation Administration. Chemicals entered commerce largely without pre-market review until the Toxic Substances Control Act of 1976, which proved too weak to pull harmful products off the market. The internet developed without a federal framework for platform liability or content moderation for its first two decades. In each case, the leading technology country deferred to industry until a crisis made deferral untenable. AI may follow the same path. Or the voluntary order may return in a form that satisfies the industry's concerns.
What happens next is unclear. Trump has not abandoned the order permanently. The White House has indicated a revised version could return. What to watch is whether any revised framework preserves any form of pre-release review, or whether the industry calls that a win and moves on. The White House also gave OpenAI explicit backing to push AI regulations at the state level, Semafor reported, a strategy that could produce a patchwork of state rules rather than a single federal standard. That approach is the inverse of the 90-day review window: instead of the federal government tracking frontier AI, state governments would set their own rules and the labs would navigate them.
Washington last year issued a separate order directing federal agencies to identify state laws that obstruct national AI policy, the White House website shows. That order remains in effect. The administration appears to be operating on two tracks simultaneously: loosening federal AI oversight while pushing back against states that want to impose their own.