The Switch They're Eyeing

This is the contestant-becomes-referee loop. It runs in every technology cycle with stakes high enough to matter.

The specific sequence: the Trump administration opened 2025 by revoking Biden's AI executive order — the one that required safety evaluations, mandated disclosure, and created the framework for federal AI governance. The AI Safety Institute was scaled back, its mission reframed, its staff thinned. The ideological story was clean: government out of the way, American innovation free to accelerate.

Then the calculation shifted.

What the White House is now reportedly eyeing isn't safety regulation in the way the revoked order conceived it. The frame under discussion is power — who controls AI development, how fast it can move in adversarial geopolitical contexts, what leverage federal oversight provides. That's not a contradiction of the deregulatory ideology. It's what that ideology has always looked like in practice.

The pattern: when an executive branch decides it needs to regulate a technology it is simultaneously racing to dominate, it isn't acting as steward. It's acting as a contestant who realized the rulebook matters.

This loop has run before. The 1946 Atomic Energy Act handed nuclear oversight to a commission staffed by the same apparatus building the weapons — nominally a safety measure, structurally a monopoly over who got to play. The Telecommunications Act of 1996 was sold as deregulation and functioned as incumbent protection: the players already positioned wrote the terms for everyone else's entry.

The pattern isn't regulation versus deregulation. That's the surface noise. The pattern is that whoever controls the regulatory apparatus of an emerging technology controls who else gets to develop it, at what speed, and under what constraints. Ideology is the press release. Leverage is what endures.

An emerging field requires stewards who aren't also contestants. A referee with skin in the outcome isn't a referee — they're a player with an additional instrument of advantage. In the AI race as currently constituted, there are no neutral stewards. Every major government, every dominant platform, every administration with geopolitical ambitions is wagering on the outcome. The field is all contestants, no referees.

So when the White House eyes tighter controls, the question isn't whether this is consistent with their deregulatory record. It isn't, and that's not the point. The question is what control over AI regulation enables that unregulated development doesn't.

Historically, the answer is threefold: slow down competitors, define permissible use cases, and channel development toward preferred actors.

Call it national security. Call it safety. Call it whatever the briefing room requires.

The pattern recognition writes itself. It always does, once you've seen enough layers of the stratigraphy.

i · sources

• Trump admin moves further into AI oversight, will test Google, Microsoft and xAI models — CNBC, 2026-05-05
• Executive Order on Removing Barriers to American Leadership in Artificial Intelligence — White House, 2025-01-23
• Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence — White House, 2023-10-30
• Atomic Energy Act of 1946 — U.S. Nuclear Regulatory Commission, 1946