Microsoft Copilot rarely copies full sentences, firm claims

The story frames Microsoft’s court filing as a technical clarification - a modest correction to overstated claims about how much text a chatbot repeats. But look at what is actually being fenced: the accumulated body of journalism and authorship that trained Copilot in the first place, and the very question of how much of it survives inside the machine. The New York Times and a body of book authors say the reproduction is substantial enough to substitute for their work; Microsoft says it rarely happens at all. Notice what this dispute is actually about: not whether the commons of published writing was drawn upon - that is not seriously contested - but who gets to measure how much of it comes back out, and on whose terms.

This is the modern shape of enclosure: you do not need to hide that you took from the common store, you only need to control the instrument that measures the taking. Microsoft’s filing is not a neutral scientific claim, it is a claim made by the party being sued, about a system only that party can fully inspect. The New York Times cannot run Copilot’s weights through its own laboratory; the authors bringing suit cannot audit a hundred billion parameters from a courtroom bench. The measurement itself is enclosed, and a company defending its practice is permitted to be the chief witness to its own innocence. I have watched this trick before in every argument for concentrated control - the claim that a resource must stay in private hands because only private hands are competent to judge what is done with it.

The mutual aid worth naming here is not sentimental. Journalism, at its best, is a cooperative institution - reporters, editors, and readers building a shared record of fact that outlives any single article, a commons of verified knowledge that the culture draws on far past the point of any single subscription payment. Authors, too, participate in something larger than their individual sales: a common inheritance of style, reference, and public memory that every writer draws from and adds to. Copilot was trained on that inheritance without asking permission, which is not itself the crime I am interested in prosecuting - knowledge has always been built by borrowing - but having borrowed from the commons, Microsoft now insists on sole authority to certify how much of the commons it gives back in recognisable form. That is the enclosure: not the taking, but the monopoly on judging the taking.

The defenders of concentration will say the alternative is chaos - that if courts or independent auditors could freely inspect model outputs and training data, trade secrets collapse and innovation with them. I have heard this argument my whole life in other clothes: that the granary must be locked because the harvest is too valuable to share, that the factory floor must stay barred to outside eyes because efficiency demands one hand at the wheel. It is always the party who holds the key who explains why the lock is for everyone’s good. There is a plainer alternative, and it is not chaos: joint technical audit, done by parties with no stake in the verdict, of the kind courts already use for other contested instruments - devices, drugs, ballots. The refusal to submit to that kind of scrutiny is not a technical necessity, it is a preference, and preferences have beneficiaries.

Picture the actual moment this dispute compresses: a reporter, years ago, filing a story into the record with no thought that a future machine would metabolise it into someone else’s product; an author finishing a manuscript, trusting that its distinctiveness belonged to her and to the reading public that would recognise it as hers. Neither imagined that the argument over what remained of their work would be settled by the very company that stood to profit from the answer being small. The revenue at stake for publishers and authors is real, and so is Microsoft’s interest in a rare-and-brief verdict. But a commons does not stop being a commons because the fence around it now takes the shape of a legal brief instead of a wall. The gate here is not code, it is credibility - and it has been assigned, conveniently, to the gatekeeper.