The world's largest scientific pre-publication server, arXiv, has limited the number of materials a single author can submit to the platform since October 2026. The reason is a sharp increase in the flow of applications, which, according to the administration, is partly related to the proliferation of advanced AI tools. Moderators have encountered an increase in low-quality, overly specialized, and AI-assisted work that takes too much time to process. A single author can now submit no more than two articles to arXiv per calendar month. At the same time, he can have no more than three active bids. Restrictions apply to all authors and all subject categories. The number of applications has doubled in two years The scale of the problem is clearly visible from the arXiv statistics. In September 2016, the platform received 9,869 applications. In September 2024, their number increased to 20,569, and in September 2026, it reached 40,363. The flow of applications has almost doubled in two years. In September 2026 alone, because of this, employees and volunteer moderators received almost 9 thousand applications to the support service. The increase in the number of materials for arXiv has become a problem not only because of the volume of the texts themselves. Each application takes time to verify, and moderators are limited. That's why the new system limits the number of submitted works, not published ones. If an article is rejected, it still counts towards the monthly limit. the work has already taken the time of the moderators. AB has accelerated the production of articles, and also their flow arXiv admits that modern AI tools can speed up individual stages of scientific work. However, at the same time, they have reduced the threshold of mass creation and sending of scientific texts. Moderators note an increase in several types of problematic materials at once. Among them are "thin" articles devoted to extremely narrow questions, the so-called "salami-articles", when the results of one study are artificially divided into several publications, as well as texts created using AB. The problem is particularly acute among the small group of authors. According to arXiv, a relatively small number of people submit a large amount of low-quality work and thus take up a disproportionate amount of the moderators' time. "This is unfair to researchers who continue to present quality works," the platform said in a statement. At the same time, arXiv does not impose a general ban on the use of AB in the preparation of scientific publications. The problem for the platform is not the fact of using such tools, but the sharp increase in the number of applications, among which there have been more materials that do not meet its standards. How does the new limit work? In addition to the limit of two applications per month, there is also a limit of three active applications at the same time. If a job was submitted in the previous month and is still under review, it does not reduce the new monthly limit in the following month. However, it still counts towards the limit of three active bids. There is also an exception for removed materials. If the author deletes the article before it is published, it does not count towards the monthly or the total limit of active applications. Thus, arXiv tries not only to reduce the overall flow of material, but also to redistribute the limited time of moderators so that it is not spent primarily on a small group of the most active authors. What will change for scientists? arXiv remains one of the most important tools for the rapid dissemination of scientific results. Unlike traditional journals, prepublications allow researchers to publish work before the lengthy peer review process is completed. The platform has existed since 1991 and today contains more than 3 million works on physics, mathematics, informatics, biology, statistics, economics and other fields. Therefore, the new limit may be a noticeable change for researchers who regularly publish a large number of results. However, the main goal of the policy, judging by the arXiv explanation, is not to limit scientific activity as such, but to try to adapt to the new reality, where text creation has become significantly cheaper and faster. This creates a paradox for the scientific infrastructure. AI can speed up the research and publication process itself, but at the same time it can multiply the amount of material that people need to review. As a result, automating one part of the scientific process creates an additional burden on the other part, the part where human expert judgment is still necessary.