What Changed

The Trump administration is actively weighing an executive order to create a working group of government officials and industry executives tasked with reviewing new AI models before public release. The review would reportedly involve the NSA, the White House Office of the National Cyber Director, and the Director of National Intelligence, according to TechPolicy.Press and The New York Times.

An administration that spent 2025 dismantling Biden-era AI oversight frameworks is now considering a pre-release vetting process — a significant reversal after months of deregulation efforts.

Why It Happened

The backstory reveals internal conflict within the Trump administration. White House AI czar David Sacks spent 2025 attempting to strip state-level AI regulations out of must-pass federal funding bills. Trump legal adviser Mike Davis blocked him twice. Their confrontation escalated to Vice President JD Vance's office in November.

Davis accused Sacks of bypassing Congress and imposing AI policy without guardrails. Sacks said he was following Trump's orders to accelerate AI development. Vance told them to resolve their differences.

In December 2025, Trump signed an executive order aimed at blocking state AI regulations and building a "single national framework," according to CNN. That order already faces expected legal challenges.

Now the administration is considering actual pre-release review. Pressure from 60-plus MAGA allies who sent Trump a letter demanding AI vetting before deployment appears to have shifted the administration's approach, according to Axios.

The Conflict of Interest

The proposed working group would be co-designed with the companies being reviewed, according to TechPolicy.Press analyst Emma Hatheway. The structure creates a fundamental problem: the industry shapes the process that evaluates itself.

Nearly 80 percent of global AI computing power is privately owned, according to TechPolicy.Press. Almost 70 percent of new AI PhDs go directly to private sector jobs — pulled away from universities by compensation packages no academic institution can match. The researchers who could evaluate whether a model poses risks mostly work for the labs building those models.

Hatheway points to Anthropic's recent decision to withhold a model called Mythos due to cybersecurity risks as illustrative. Anthropic made that call unilaterally — one company, a handful of executives, a private decision affecting millions of potential users. That remains how AI safety decisions work in America today.

A government panel staffed by intelligence officials and industry representatives doesn't resolve the independence problem. It adds federal authority to a process where the regulated parties help design their own oversight.

What the Coverage Misses

CNN frames this as a "MAGA divide" story — tech optimists versus working-class skeptics. The framing captures real tensions but obscures the central question: who controls the most consequential technology in human history, and whether any independent institution has the capacity to evaluate it objectively.

Currently, no independent mechanism exists. A working group co-staffed by industry executives is not an answer to that problem.

Axios reported that 60-plus MAGA allies want Trump to vet AI before release. The critical detail is what vetting actually entails. A review process with no enforcement mechanism, staffed by people with financial interests in approval, functions as political theater rather than genuine safety oversight.

Sacks' Argument

Sacks is correct that America needs to move quickly on AI to maintain technological advantage over China. That argument has legitimate national security dimensions, not just industry appeal. China continues advancing AI capabilities.

But acceleration without independent safety mechanisms creates different risks. A catastrophic deployment would damage U.S. credibility internationally and derail domestic AI policy for years.

The administration wants both deregulation and a safety framework. A process controlled by the regulated industry accomplishes neither. It removes state and federal oversight while claiming to establish safeguards.

The Stakes

The models under review will determine what Americans see online, what medical diagnoses get flagged, what credit scores get calculated, and what jobs face automation. The vetting process is being designed by the companies profiting from those decisions.

Trump's team moved on AI policy. The direction and speed of change, however, remain inadequate to address the underlying conflicts. A process built on industry input is fundamentally compromised before it begins.