Schneier Calls For US Nationalization Of AI Giants

The story celebrates that OpenAI and Anthropic were built - the funding rounds, the model releases, the talk of frontier capability arriving on schedule. But a made thing does not stop where its maker’s attention stops; it goes on acting in a world no lab contains. Bruce Schneier and Nathan E Sanders, writing on this proposal to nationalize such companies should the markets reject them, have skipped past the launch to ask the only question that lasts: who is answerable for what these systems do after release, and what did their makers fail to imagine when they imagined only shareholders?

The proposal is worth taking seriously precisely because it does not pretend the market is a neutral judge of safety. Markets price capability and growth; they do not price the years-long tail of consequence that begins once a model is loosed into hospitals, courtrooms, classrooms, and the private correspondence of the lonely. Schneier and Sanders are asking what happens if the ledger the market keeps and the ledger the public needs turn out to be different documents entirely - and proposing that the US government hold the second one when private capital abandons the task. This is not doom-mongering. It is a plan for who inherits the debt when the original debtor walks away.

But notice what the proposal does not solve, and here the strongest objection deserves its due: nationalizing OpenAI or Anthropic does not transfer the wisdom to be accountable, only the legal liability. A government agency can inherit servers and staff; it cannot inherit the tacit knowledge of why a model behaves as it does, knowledge that lives disproportionately in the heads of engineers who may simply leave for Google or Meta, companies conveniently absent from any nationalization scheme because their AI divisions sit inside profitable, diversified empires that markets will never “reject” in the way a pure-play lab could be rejected. The proposal’s blind spot is not too little ambition but too narrow a target: it treats the failure mode as insolvency, when the more likely failure mode is a company that remains profitable while quietly exporting harms it has no incentive to measure.

This is the gap Schneier and Sanders are groping toward without quite naming it: the distance between the capability to build a system and the wisdom to be answerable for it does not close when ownership changes hands. It closes, if it closes at all, when someone is structurally obligated to keep watching after the applause ends. A market obligates a company to keep watching only its revenue. A government, nationalizing a failed AI firm as a bank in 2008, would obligate itself to keep watching solvency. Neither ledger asks what the system is doing to the teenager who talks to it every night, or the caseworker who now defers to its judgment, or the small claims court quietly automating its docket. The debt these companies owe was never fiscal at root; it was owed to the users the model was never tested against, and no restructuring of the balance sheet discharges that.

What the proposal gets right, and what makes it more than a thought experiment, is the refusal of the convenient answer: that if the market fails, the harm simply stops, because the company folds and the model goes dark. Systems like this do not go dark quietly. Weights get open-sourced, forked, absorbed into products nobody remembers came from OpenAI or Anthropic at all, running on in a world their original makers stopped watching the day the money ran out. The nationalization question, properly asked, is not “who owns the company” but “who is legally compelled to keep asking what this is doing,” long after the founders have moved to their next venture and the story has moved to its next launch. Schneier and Sanders have located the right crisis. They have not yet found the mechanism that survives the moment the government, too, gets bored of watching.