Trump Pulls AI Safety Executive Order at the Last Minute After Calls From Musk, Zuckerberg, and Sacks

President Donald Trump abruptly postponed the signing of a long-awaited AI safety executive order on Thursday, May 22, leaving frontier-lab executives in the air on planes to Washington as the ceremony was scrubbed from the schedule. The reversal came after a coordinated series of late-night and early-morning calls to the president from xAI CEO Elon Musk, Meta CEO Mark Zuckerberg, and former White House AI and crypto czar David Sacks, according to reporting from Axios, Fortune, and the Washington Post.

The draft order would have established a voluntary mechanism for AI developers to submit frontier models to federal agencies up to 90 days before public release, with the Treasury Department — rather than CISA or NIST — taking the unusual lead on identifying and patching security vulnerabilities. The text explicitly ruled out mandatory licensing or pre-approval, but Sacks reportedly told Trump on Thursday morning that the voluntary structure would function as a "de facto licensing regime" and warned that a future administration could easily flip the same machinery into a mandatory pre-clearance system.

Speaking to reporters later that morning, Trump confirmed the postponement on competitiveness grounds: "I didn't like certain aspects of it. I think it gets in the way of — we're leading China. We're leading everybody, and I didn't want to do anything that's going to get in the way of that." The framing matched arguments long pushed by Musk and Sacks, both of whom have publicly argued that any U.S. AI safety regime hands a structural advantage to Beijing-backed labs that face no equivalent pre-release scrutiny.

The reversal exposes a striking split inside the president's own coalition. Anthropic and OpenAI had both publicly indicated they would cooperate with the 90-day review window — Anthropic co-founder Jack Clark used an Oxford lecture the same week to call for exactly the kind of pre-deployment evaluations the order would have formalized. On the MAGA side, Steve Bannon and more than 60 conservative figures had signed an open letter urging Trump to ship the order, and a Future of Life Institute poll found that 79% of Republican voters supported government testing of frontier AI models before release.

What happens next is unclear. The White House Office of the National Cyber Director said it is developing additional AI security initiatives outside the executive-order track, while Chief of Staff Susie Wiles publicly slapped down a separate FDA-style licensing proposal from National Economic Council Director Kevin Hassett, posting that the president was not "in the business of picking winners and losers." For the frontier labs, the practical effect is that the United States has now spent two consecutive administrations announcing — and then walking back — pre-release safety review regimes, leaving voluntary commitments from Anthropic, OpenAI, and Google DeepMind as the only operative U.S. safety floor.