Schneier Calls For US Nationalization Of AI Giants

The proposal arrived on a timeline that assumed institutions capable of processing it at a speed they last achieved in 1933, when the state absorbed the banking system because the banking system had absorbed everything else first. Bruce Schneier and Nathan E. Sanders, writing in the register of policy prescription rather than prophecy, suggest that the government nationalize OpenAI or Anthropic should the market decline to discipline them. The suggestion is rational. It is also roughly the speed of a nineteenth-century harbor pilot proposing to steer a vessel that has already left the visible ocean.

The dynamo here is not the artificial intelligence itself but the compounding of capital and inference speed behind it - the thing that lets a company like OpenAI or Anthropic double its relevant capability in an interval shorter than any legislature’s committee calendar. Schneier and Sanders propose a nineteenth-century remedy - sovereign seizure of a failed enterprise - for a problem whose defining feature is that it does not fail in the old sense. Railroads went bankrupt visibly, with unpaid bondholders and idle track. A frontier AI lab that “fails” in market terms may still be shipping systems that reorganize labor markets, or persuasion, or warfare, at a rate the failure metric never registers. Nationalization presumes a corpse to inherit. The dynamo produces no corpse; it produces continuous operation at diminishing accountability.

Note what the proposal leaves out, and note it is not an oversight but a structural fact: Google and Meta sit outside its logic entirely, not because they are less powerful but because they are not organized as single-product AI ventures a state could plausibly absorb in one gesture. The remedy is shaped to fit two companies precisely because those two companies are shaped like nineteenth-century monopolies - legible, seizable, nameable. The actual diffusion of the dynamo runs through infrastructure Schneier and Sanders’ framework cannot reach: cloud compute, chip fabrication, the training runs quietly proceeding inside firms whose primary business is not called intelligence at all. This is the education gap in its purest form. The proposal is calibrated to the shape of the crisis of 1911 - the trust-busting shape, the shape of Standard Oil broken into pieces on a courtroom’s schedule - applied to a crisis whose shape is closer to weather than to monopoly.

The entropy is worth measuring precisely because it is not chaos but its opposite: too much order, arriving too fast, concentrated in institutions built to be liquid, changeable, unaccountable to any electoral clock. A senator conducting a hearing on Anthropic’s capabilities in 2026 is conducting the hearing his predecessors were equipped for - the hearing about railroad rates, about oil pipelines, about the telephone monopoly - updated only in vocabulary. The apparatus of nationalization assumes a state with the technical comprehension to run what it seizes, and no such state exists yet; the relevant expertise resides, disproportionately, inside the very companies proposed for seizure. One does not solve a comprehension gap by transferring ownership across it.

What the proposal gets right, and what a purely skeptical reading would miss, is the diagnosis beneath the remedy: that the market alone cannot be trusted to align an artifact whose value proposition is inseparable from the possibility of catastrophic misuse. This is true, and true in a way that recalls the earlier moment when the Virgin’s world - the world of settled meaning, moral confidence, comprehensible scale - met an electrical grid it could not narrate. Schneier and Sanders are not wrong that the market’s incentives run toward speed and away from caution. They are proposing the correct diagnosis and the wrong century’s remedy, which is the recurring shape of every serious proposal in this field: accurate about the danger, decades behind on the mechanism.

The measurement, then, is this: the interval between the emergence of a genuinely dynamo-scale technology and the emergence of an institutional vocabulary adequate to govern it has not shortened with practice, as one might hope skill would compound. It has lengthened, because each new dynamo arrives inside institutions still processing the last one’s paperwork. Somewhere in a data center a training run completes on schedule. Somewhere in Washington a hearing is scheduled for next quarter. The gap between those two clocks is not a policy failure. It is the policy.