President Trump shelved a long-deliberated executive order on AI risk after an 11th-hour phone call from David Sacks, the venture capitalist and current administration AI advisor, the Wall Street Journal reported on May 22. Sacks argued that the order, which would have imposed mandatory regulations on frontier AI labs, would slow the U.S. industry in its competition with Chinese AI development. Trump responded that he shared concerns about China, expressed worry about hindering AI investment, postponed the signing, and later told reporters he would not sign the order at all.
The reporting frames Sacks as the decisive influence on a policy direction that had been telegraphed for weeks as the administration’s signature AI safety move. The order, in its draft form, would have introduced binding rules on training-compute thresholds, capability disclosure, and incident reporting for the largest frontier labs. Industry trade associations had been lobbying against versions of these requirements since the draft surfaced in March. Sacks’s call closed the loop.
Sacks is not a neutral figure in this debate. He was an early member of the PayPal Mafia, has been a vocal Silicon Valley supporter of the Trump administration, and currently holds an advisory role focused on AI and Web3 policy. He has spent the past two years arguing publicly that any binding regulation on U.S. frontier labs is a strategic gift to China. The Wall Street Journal’s reporting positions his call as singular, but it is more accurate to describe him as the formal articulator of a coordinated industry position that has been building since the EU AI Act took effect.
The structural reality is that the AI industry has been lobbying with unusual unity against any binding U.S. regulation since 2024. OpenAI, Anthropic, Google, Meta, and Microsoft have publicly differed on capability scaling, safety posture, and product strategy, but on the specific question of whether the U.S. government should impose mandatory rules on training and deployment, they have moved as a bloc. Sacks’s call is a synecdoche for that coordinated position, not the whole story. The Wall Street Journal’s reporting frames him as the influencer; the institutional story is that the entire frontier-lab industry was pushing the same outcome.
The result clarifies the U.S. regulatory landscape for the next 12 to 24 months. There will be no binding federal framework for frontier AI labs in the near term. Voluntary commitments, of the kind Anthropic, OpenAI, and others signed under the Biden administration, will be the dominant compliance mechanism at the federal level. State-level rules, including California’s SB 1047-class proposals, will continue to fragment the landscape, but federal preemption is not imminent in either direction.
For frontier labs, this is the policy outcome they have been engineering since 2023. It also concentrates the responsibility for safety practices entirely on internal lab governance, since no external enforcement mechanism will exist. Anthropic’s exploit-evals publication this week, and the broader pattern of labs publishing self-assessments of dangerous capabilities, becomes the de facto evidentiary record that regulators do not have. Whether that is sufficient is a political question that the federal government has now declined to answer.
For policy and government-relations teams at AI-deploying enterprises, the practical takeaway is to plan procurement and compliance against a 2027 horizon assuming no binding federal AI law and continued state-level patchwork. EU AI Act compliance remains the dominant external requirement for any U.S. lab serving European customers; that constraint has not changed and is now relatively more important because no equivalent U.S. framework will exist to displace it.
Reported by the Wall Street Journal on 2026-05-22.
