Publishers are restricting the Internet Archive’s access to their journalism over concerns about AI training, the archive’s founder argues in The Atlantic. The consequence, he says, is a thinner public record that journalists and researchers rely on to check what once appeared online.
Watch Desk analysis
What happened
The Internet Archive says news organisations including The New York Times, The Guardian and USA Today have placed barriers in the way of its web-preservation work, citing unauthorised use of archived material by AI companies. The Atlantic essay argues that blocking a nonprofit archive will not stop powerful companies from gathering training data, but will make historical reporting harder.
The essay cites a 2024 Pew Research Center study that tracked one million webpages over a decade: a quarter were no longer accessible by the end of the study. The Internet Archive’s analysis found that 16% of vanished URLs in the study’s data could be accessed through the Wayback Machine. The essay also says Harvard’s Nieman Journalism Lab had identified more than 340 local-news sites limiting the archive’s ability to preserve their journalism.
Why it matters
The dispute is not simply about whether AI companies should be allowed to train on journalism. It is also about who can preserve a record of what publishers, public bodies and other organisations put online, and who can consult that record later. The essay points to journalists using archived pages to recover deleted material and compare changes over time.
Those interests need not be treated as identical. Publishers can challenge commercial use of their work while still recognising the public value of preservation. The awkward question is how to protect both without making the historical record collateral damage in the fight over AI training.
Our read
The essay makes a persuasive case that preservation deserves its own place in the debate, rather than being bundled into a single yes-or-no argument about AI scraping. But the archive’s account is one side of a live dispute: publishers’ concerns about control, access and commercial use deserve to be addressed, not waved away. A useful next step would be clearer rules distinguishing preservation from commercial AI use.
What to watch
- Whether publishers and the Internet Archive agree on preservation access that does not enable unrestricted commercial reuse.
- How blocking and removal policies affect journalists’ ability to check historical versions of online material.
- Whether policymakers or courts draw a clearer line between archiving journalism and using it to train AI.
Discussion spark: Should publishers be able to block nonprofit web archives to limit AI training, even if that makes old reporting and public records harder to verify?
Sources and evidence
- The Internet Needs a Memory – The Atlantic (19 October 2026, 16:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.