Imagine a library where, every month, twice as many books arrive as two years ago. Those sorting them are largely volunteers. One day, they put up a sign at the entrance: two books per person per month, not one more.
That is exactly what has just happened at the largest online library in scientific research.
What is arXiv?
arXiv is a free site where researchers publish their work before it even appears in a journal. Physics, mathematics, computer science: this is where science is shared first. For artificial intelligence, it is even the main venue: almost every discovery in the field appears there first.
It is not a scientific journal. Articles are not peer-reviewed in depth there. Moderators, often volunteers, simply check that they look like genuine research before putting them online.
The figure that triggered everything
In September 2026, arXiv received 40,363 articles. A record. That is twice as many as two years ago, and four times as many as ten years ago, when fewer than 10,000 arrived per month. In the category devoted to AI, submissions have increased more than sixfold in two years.
These submissions also generated nearly 9,000 support requests. For a team that relies largely on volunteers, it was untenable.
The new rule
Since 1 October, each author can submit no more than two articles per month, and have a maximum of three articles pending at the same time. Rejected articles count too: a rejected text uses up one of the two slots for the month.
arXiv explains its decision as follows: a small proportion of authors submit a large number of low-quality articles, and monopolise moderators' time. It is unfair to all those who submit serious work and wait their turn.
The underlying reasoning is even more interesting. Before AI, the speed at which a researcher could write set a natural limit: nobody drafted twenty serious articles a month. AI removed that limit. arXiv is restoring it through the rules.
arXiv was born in 1991, at the American laboratory in Los Alamos, on the initiative of physicist Paul Ginsparg. At first, it was a simple email list so that physicists could exchange their work without waiting months for publication in a journal. Thirty-five years later, this makeshift service hosts millions of articles and serves as a gateway to a good part of world science. It still runs, in large part, thanks to the goodwill of volunteers.
A decision that is being debated
Not everyone is applauding it. An honest and highly productive researcher, working with several teams, may legitimately sign more than two articles a month. They will now have to wait.
And because rejections count, some fear that authors will no longer dare to submit original but unusual work, for fear of losing a slot. For critics, it is an organised shortage: the quantity of science entering this great public registry now depends on moderators' time, and no longer on the science being done.
What we take away
This story goes beyond arXiv. When producing a text costs almost nothing, the scarce resource is no longer writing. It is the attention of those who have to read, sort and verify.
We wrote it in September in our reflection on slowness: when production outpaces verification, something has to give. Here, it is access to the registry of science that has been rationed. It probably will not be the last.