When the monopolist asks for his own leash

The Reason piece puts the matter plainly: Dario Amodei, chief executive of Anthropic, is calling for coordinated international limits on artificial-intelligence development at the precise moment his company is preparing to go public. One need not be a cynic to notice the timing. One need only have spent any time studying what I called, in The Wealth of Nations, the "mean rapacity" and "monopolizing spirit" of merchants who, having climbed a considerable height, petition the sovereign to pull up the ladder behind them.

I do not say that Mr. Amodei's fears about powerful AI are insincere. A man may hold a genuine alarm and a self-serving business interest simultaneously — indeed, that is the ordinary condition of the merchant mind. The question a political economist must ask is not what the executive feels, but what the proposed regulation does: who bears its cost, and who reaps its benefit. Compliance with elaborate international licensing schemes costs more, as a share of revenue, for a small entrant than for an established firm with lawyers, lobbyists, and the ear of regulators already in place. The barrier that looks like a safety rail functions, in practice, as a tollgate.

This is the pattern I traced through the mercantilist legislation of my own century. The great chartered companies — the East India Company foremost among them — did not oppose all regulation. They authored it, then presented it to Parliament as a public good. The public, who bore the higher prices, rarely appeared in those proceedings. I would ask any reader of the Reason account to notice who is absent from Amodei's coalition: the smaller laboratories, the open-source researchers, the consumers of future AI services. They will pay the compliance cost and receive no seat at the drafting table.

Now, I am not indifferent to the genuine public-good argument. There are matters — defense, justice, infrastructure, and education — where private calculation cannot supply what the public requires, and the sovereign must act. If artificial intelligence poses catastrophic risks that no individual firm has sufficient incentive to prevent, that is precisely the kind of externality that calls for public authority. I grant the argument its due weight. But the institutional form matters enormously. A regulatory body designed with the participation of incumbents, and administered by officials who rotate between the regulator and the regulated, is not a public good. It is a private toll dressed in public clothing.

The honest version of Mr. Amodei's proposal would begin not with an international treaty drafted among the largest players, but with transparency: open publication of safety findings, independent auditing bodies with no financial tie to the firms they examine, and — crucially — rules that apply equally to the firm seeking the regulation and to its smallest competitor. Where I argued, in The Wealth of Nations, that the solution to mercantile capture was not to destroy commerce but to make its terms public and its courts honest, I would argue here that the solution is not to ban AI development but to build the institutional framework that makes its practice accountable. Sunlight, as it were, is the better disinfectant — and it costs the honest actor nothing he should wish to hide.