The Internet Archive published “An update on Wayback Machine access” on Tuesday, September 15. By Wednesday morning it was sitting at the top of Hacker News with 632 points and 339 comments, which is the internet’s way of announcing that a website most people consult twice a year has become load-bearing. The commenters are right that losing the Wayback Machine would be bad. They are wrong about what this fight is actually about.

The Blocking Isn’t Aimed at the AI Labs

Start with who is doing it. Nieman Journalism Lab counted more than 340 local news outlets limiting the Archive’s access to their journalism as of May 20. Forbes reported in April that 87 percent of the blocking sites are owned by USA Today Co. — the chain formerly called Gannett, with more than 200 papers — and that The New York Times went beyond the usual robots.txt conventions at the end of 2025, hard-blocking Archive crawlers. The Atlantic has been working with Cloudflare on it; Condé Nast has rolled the policy across Vogue, The New Yorker, Wired and the rest. Deutsche Welle covered the whole thing this month under the headline “Digital memory at stake.”

Now notice the timing problem. If this were about stopping AI training, every one of these moves is years too late. The crawls happened. The snapshots are in the datasets. You cannot un-train a model by editing a text file in 2026 — and the publishers’ own licensing people know it, because they are simultaneously selling the 2019 crawl to the labs they claim to be fencing out. The block is not a defense against machine learning. It is a demonstration of control, staged for a counterparty that isn’t the Archive.

What They Want Back Is the Erasure

A newspaper’s back catalog is its largest uninsured liability. It holds editorials that aged into embarrassments, sentences that were quietly softened after a phone call, the names of people who should never have been printed, and corrections appended at 4 p.m. that most readers never saw. The Wayback Machine freezes every draft, including the bad ones. That is the entire point of it, and it is the reason the blocking is concentrated among institutions with a long paper trail and a short institutional memory.

Note what the Archive never claimed to be. It is not a library in the sense the mission statement suggests, and it never preserved journalism out of charity. It preserved it because a public record nobody can check is not a record. That distinction is doing all the work here, and nobody in this fight wants to say it out loud.

A Paper Trail Only Works If Somebody Keeps It

The use case the coverage skips is the boring one. “We can only prove a product page from 2011 ever existed because somebody archived it,” a document-retrieval paralegal in Phoenix told me, between depositions. “Take that away and the party with the bigger legal budget wins by default.”

That is what the Archive actually does, and it has very little to do with posterity. It makes institutions’ past statements binding on them — a hospital’s posted charity-care policy, a county’s water advisory, a chemical company’s deleted assurance about a solvent, a bank’s fee schedule from before the fee moved. All of it checkable, which is to say actionable. That is a deeply conservative function: it is the paper trail that lets an ordinary person hold a bureaucracy to what it said. It is precisely the function being dismantled, by a coalition of institutions that would prefer to retain the option of having said something else.

Which leaves the right in a spot it has not acknowledged. Property rights and accountability are now in open conflict, and the reflex to defend the former has a cost the movement hasn’t priced. You can have the doctrine that an archive is someone’s property to withhold, or you can have the doctrine that institutions are bound by what they published. Pick one. The Gannett papers have already picked.

The Small Papers Are Paying for the Big Ones’ Leverage

The New York Times can afford a hard block. It has a brand, a subscription base, and licensing leverage worth arguing over. The 340-odd local outlets mostly have none of that. A Gannett daily’s 1994 school board investigation has no AI licensing value and no readers except the ones who arrive through a search result that now resolves to a dead link. Those papers are burning the only asset that made them citable in order to prove, to a buyer who isn’t watching, that they own it.

The Archive, for its part, co-hosted the National Summit on Local News Preservation on June 17 in National Harbor, Maryland, with the Poynter Institute and Investigative Reporters and Editors — a room full of people trying to keep local news alive, while a third of the industry deleted its own back catalog from the last index that pointed at it. And the whole arrangement rests on infrastructure frail enough that an August power failure took the site offline. When the thing you’re defending is both load-bearing and uninsured, the fight over who may copy it is the wrong fight.

The 339 Hacker News commenters are mourning a library. What they’re actually about to lose is a ledger — the one place where what an institution said in public stayed on the books. Everyone claims to want that, right up until the sentence on file is theirs.

Sources