The Trump administration is tangled in a fierce battle over how to regulate artificial intelligence. The president pulled the plug on a key executive order just hours before its signing.

This order aimed to create a voluntary framework for AI companies to give the government early access to their most advanced models. The goal was to evaluate cybersecurity risks before public release. But Trump worried it would undercut America’s lead over China.

Inside the White House, factions clash. One camp, led by former AI czar David Sacks, pushes for minimal regulation. Sacks convinced Trump the order would throttle innovation and create bureaucratic roadblocks. He won the last-minute cancellation.

Opposing them are officials like Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross. They back a middle ground—voluntary disclosure allowing early government review without heavy-handed mandates. Their focus: balancing national security with industry growth.

Commerce and defense departments are mostly sidelined. The Pentagon wants early access to AI tech but avoids full regulatory engagement. Meanwhile, the Commerce Department’s existing AI testing program quietly handles some model reviews without formal rules.

The administration also seeks a nationwide AI law to prevent states from imposing their own competing rules. This would protect business interests and avoid a patchwork of conflicting regulations. The plan includes child safety controls, anti-scam measures, and energy policy adjustments for data centers.

States like New York and California oppose federal preemption. They argue local oversight is critical to address AI’s social risks. The White House counters that a fragmented approach threatens U.S. leadership in the global AI race.

Trump’s reluctance to regulate AI stems from a desire to keep America dominant over China. Yet cybersecurity threats from new models—like Anthropic’s Mythos, which has uncovered thousands of software vulnerabilities—push the administration toward caution.

The executive order also tasks the Justice Department with cracking down on AI-enabled crimes, focusing on misuse of autonomous AI systems. This signals growing concern over AI’s capacity for both innovation and exploitation.

Industry reactions are mixed. Some executives find the regulatory process chaotic and unclear. They worry about handing over proprietary models months before release. Others welcome a framework that addresses cyber risks without imposing licensing requirements.

The cancellation of the executive order has left the administration back at square one. Staffers say the policy is “still to play for,” hinging on whether Trump will reverse course and approve a revised version. Until then, the AI sector faces uncertainty amid growing geopolitical and security stakes.

This internal friction reflects a broader tension: how to govern a technology that promises enormous economic growth but carries unprecedented risks. The administration’s struggle highlights the difficulty of balancing innovation with oversight in a critical emerging field.

Without clear federal rules, investment decisions may slow. Companies need certainty to navigate AI’s fast-moving landscape. The U.S. risks falling behind if regulatory chaos persists or if it opts for deregulation that ignores security concerns.

For now, the White House is caught between competing visions—those who fear regulation will kill American AI leadership and those who see federal oversight as essential to protect national interests. The outcome will shape AI’s future in the U.S. and its global standing.