I think this is a bad idea.
First, let me say that the arXiv has been an enormous net good for the community. A lot of people do largely thankless work behind the scenes to keep it running, and I understand why they would want to introduce a policy like this.
Second, I would be happy to live in a world where people actually produced at most two papers a month. I don’t think I have ever produced more than two papers in a single month, and I don’t have plans to start doing so. There certainly are cases of people I respect and admire having produced more than two high-quality papers in a month on occasion, well before AI assistance, but that isn’t my main point.
The problem is that limiting arXiv submissions does not limit the production of papers. It limits when those papers become publicly available through the place we all look for them.
The arXiv serves both as a public timestamp and as a common record of new results. If I want to know whether someone has just solved a problem I am working on, I can mostly monitor the arXiv each night. That system is imperfect, but it is enormously useful.
Under this policy, we will increasingly see situations where someone posts a paper and then receives an email saying that someone else obtained the result a month earlier, but their paper is waiting for an available submission slot or has only been submitted to a journal. People will continue to circulate papers privately, both to establish priority and to share results with interested colleagues. But the rest of the community will have a less complete record of what has been done and when.
The arXiv lets the community share results and establish a public timestamp, with everyone having access to the same information at the same time. Moving more of this into private correspondence will make both priority and access to new results depend more on whom you know. That will make our community less open and democratic.
I’m all in favor of trying to change some of the behaviors that have arisen with AI. But we should do that by changing the incentives that encourage those behaviors, not by changing policies in ways that will (predictably) create other problems.