Schneier Calls For US Nationalization Of AI Giants

The story frames Bruce Schneier and Nathan E Sanders’s proposal as a safeguard - nationalize OpenAI or Anthropic if the market rejects them, so that catastrophe is averted by public hand rather than private greed. But look at what is actually being fenced: the trained weights, the accumulated corpus of human writing and code that built these models, and the governing authority over what they may say and do. That is the commons at stake, and the proposal, however well meant, is a debate about who holds the fence, not whether one stands.

Schneier and Sanders deserve credit for naming something true: markets can fail, and when they fail at the scale of frontier AI, the failure is not contained to shareholders. But their remedy assumes the only alternative to corporate enclosure is state enclosure - that the fence must exist, and the only live question is whether the hand on the gate wears a company badge or a government one. I have heard this argument before, always from people about to become the gatekeeper. It was made about grain, about land, about the telegraph. It is made now about the weights that OpenAI and Anthropic have built substantially from the shared writing, code, and conversation of millions who never signed a nationalization clause and never signed a licensing agreement either. The knowledge commons that trained these systems was never theirs to sell, and it will not become the public’s simply because Washington buys the company that fenced it.

Consider who actually gains if the US government absorbs a failing OpenAI. Not the programmer in Lagos or the researcher in Manila who contributed, unpaid and unnamed, to the datasets and open discussions that made large language models possible before anyone called them a national asset. Not the small lab experimenting with open weights, which now competes against a firm carrying the full weight of federal procurement and classification power. The beneficiaries are Google and Meta, watching a rival stumble into a form of protection neither of them enjoys, and the state itself, which acquires a lever over information flow that no ordinary regulator could dream of. Nationalization does not open the model to the many who built its raw material. It swaps one owner’s balance sheet for another owner’s mandate, and calls the swap democracy.

The pretext, as always, is safety - the claim that catastrophe can only be averted by concentrating control, never by distributing it. This is worth sitting with, because it is not stupid. A frontier model genuinely might be dangerous in ways that reward careful custody. But notice the sleight: the argument for concentration is offered only after enclosure has already occurred, after OpenAI and Anthropic have spent years walling their weights away from the public that supplied the words those weights were trained on. The choice on offer - corporate control or state control - was manufactured by the prior act of closing what had been, in its raw material, a shared thing. Nobody is proposing the third option: publish the weights, open the training process to audit by the many rather than custody by the few, and let mutual scrutiny do what neither a boardroom nor a ministry reliably does alone.

I think of the archivists who spent a decade building the open corpora that made this whole field possible, uncredited, mostly unpaid in any currency but curiosity, now watching two companies and possibly one government argue over who owns the harvest. That is the commons still being enclosed, quietly, under cover of a debate about markets and rescue.

Nationalization dressed as safety is still a fence, only with a flag on it instead of a logo. The real alternative was never on the table in this proposal, and that omission tells you more about the limits of the imagination in Washington and Silicon Valley alike than any market failure could.