The oldest trick in rivalry: make your competitor afraid of his own tools

The National Review puts it plainly: Congress needs the discipline to distinguish between a guardrail and a roadblock — and to ask, critically, who benefits from each. That question is not merely procedural. It is the question on which the industrial future of this republic will turn.

I spent the better part of my public life arguing that manufactures are not a luxury appended to an agricultural economy — they are the sinew by which a free people achieves commercial independence. The great productive technologies of any age are not neutral; they belong, in practical effect, to whoever builds the institutions, the capital, and the trained workforce to command them. Artificial intelligence is, by every inference I can draw from what has been described to me, exactly such a technology: general-purpose, compounding, and foundational to the industries that will define the next generation of national power.

China understands this. And if the National Review's lead is to be credited, China also understands something about legislative psychology: a rival nation that can be made to fear its own capacity will regulate itself into second place without a shot being fired. The fear campaign is the strategy. Every redundant compliance layer, every moratorium dressed up as precaution, every bill that answers a Chinese threat by restricting American engineers — these are not defeats China inflicts on us. They are defeats we inflict on ourselves, on China's behalf.

I will be candid about my limits here. The precise engineering of large language models, the architecture of training runs, the mechanics of inference at scale — these are beyond what I could have known, and I will not pretend otherwise. But I do not need to understand the machinery to ask the durable questions: Does this technology enlarge American productive capacity? On what terms is it deployed? Where private power concentrates without adequate check, what public institution provides the counterweight? Those are the questions my Reports on Manufactures were always asking, dressed in the language of wool and iron rather than silicon.

The answer to concentrated private power in AI is not to hobble the technology. It is to build the public architecture — standards, accountability, a vigorous federal commerce power applied with purpose — that keeps the benefits broad and the dangers bounded. A feeble response to a genuine industrial challenge is not safety. It is surrender with paperwork attached.

My recommendation is this: Congress should draw a sharp, principled line between regulations that address demonstrable, specific harms and restrictions that merely answer anxiety with prohibition. Pass the former; reject the latter. Fund the research base. Invest in the workforce. Use the federal commerce power to ensure that the gains of this technology flow into American productive capacity broadly, not merely into a handful of private balance sheets. And do not — under any circumstances — let a rival's information campaign set the terms of your industrial policy. That is not caution. That is capitulation.