The Authors Guild’s write-up of the unsealed briefs in its case against Microsoft and OpenAI hit the front page of Hacker News today — 540 points, 480 comments, most of them some version of “they knew.” The Guild’s framing is blunt: top executives understood that training on pirated books was illegal, and the documents now say so in their own words.

I haven’t read the exhibits. What I have is a party’s summary of a party’s evidence, filed by people who want a verdict. Treat it accordingly. The interesting question isn’t whether the documents are damning. It’s what kind of remedy they make possible — and whether the people celebrating this week have thought past the headline.

The Defense Was a Posture, Not a Belief

Consider what it means for a company’s general counsel to have known. If you believe fair use covers your training corpus, you don’t need to know where the books came from; the provenance is irrelevant. The knowledge only matters if you expected the argument to fail eventually — which means the public defense was never a position. It was a schedule.

That reading is uncharitable to OpenAI and Microsoft, and it should be. But it is equally uncharitable to the plaintiffs’ theory of willfulness, because it describes not two rogue companies but an entire industry that made the same calculation and put it in writing. One junior in-house lawyer at a mid-size model developer put it to me in a Slack DM with the flatness of someone who has already stopped defending it: “We priced copyright risk at zero — not because we thought we’d win, but because we thought the number would never arrive.”

That is the sentence the unsealing actually produced. Not “we stole.” “We deferred.”

Collective Knowledge Is a Settlement Machine

The plaintiffs’ narrative is that everyone knew. It is a terrific story and a mediocre litigation asset, because the more universal the knowledge, the less the case is about these two defendants. Courts don’t enjoin an industry. Congress doesn’t either. When infringement is mass, diffuse, and individually small, the American answer is a license with a collecting society attached — the thing that happened to player pianos, radio, cable retransmission, and every other technology that arrived before anyone could figure out who owned what.

Which is why the Authors Guild should be careful about how much it enjoys this week. A compulsory license pays a rate. It does not pay a novelist’s idea of justice, and it retroactively blesses the conduct being complained about. It also converts the Authors Guild from a plaintiff into a stakeholder in a rate-setting proceeding — a much duller room, with much duller numbers, and no front page.

Meanwhile the right’s preferred fix — a federal statute settling training data once and for all, so American labs aren’t litigating in fifty statehouses — becomes harder to sell the moment the record looks like this. You cannot argue that this is a nuisance suit brought by coastal publishers when the paperwork suggests the industry privately agreed with the publishers and simply declined to act on it.

What Boards and Buyers Will Reprice

The people who should be sweating are not the ones on Hacker News. They are the buyers. For three years, enterprise procurement teams and their insurers have treated training-data exposure as a known, priced risk. Unsealed documents that show the risk was understood and never quantified are the opposite of priced — they are deferred, which means the liability sits somewhere on the balance sheet of a company that thought it had bought an indemnity.

Watch what happens to every IP indemnity clause, every tech errors-and-omissions renewal, every diligence questionnaire handed to a model vendor. Watch what a board does when its outside counsel’s memo from 2023, written in the conditional, gets read aloud in a deposition. None of that requires a verdict. It requires only that the documents exist and are no longer sealed.

This is the part both sides keep missing. The plaintiffs want a finding of theft; the defendants want a finding of fair use; and what the unsealing has actually produced is a fact pattern so general that neither is likely to get one. Universal conduct produces legislative remedies. Ambiguous conduct produces settlement. Badly documented conduct produces insurers writing exclusions.

The authors may win the argument and lose the remedy. The companies may lose the argument and keep the business. That is not a satisfying ending for anyone, which is a decent sign it’s where this is going — and a good reason to read the next filing for the money, not the outrage.