The Association for Human Mathematics (AHM), the group that recently accused OpenAI of a “total disregard for norms of scientific research” after the lab’s mass release of AI-generated proofs, has grown to over a thousand members, more than 600 of whom have gone a step further and pledged to not use AI tools in their research at all.

According to the association’s public membership page, AHM now counts 1,008 members, of which 654 belong to what it calls its “AI-free caucus,” a group of mathematicians who have formally committed to not using AI models in a research capacity, including for finding proofs, exploring examples, or posing research questions.

Every AHM member, regardless of whether they join the AI-free caucus, signs onto two baseline commitments on joining: to not work with commercial AI companies by offering technical labor, mathematical knowledge, consultation or publicity, and to not publish mathematical writing, including papers, referee reports, lecture notes or books, that has been generated by AI models. The group says mathematical writing should carry human authorship, and members are expected to make their collaborators aware of these commitments as well.

The AI-free caucus commitment is optional and goes further, with members pledging to keep AI entirely out of their core mathematical work. The association updates its public member list roughly once a week, and membership is open to anyone in the mathematical research community at the graduate level or higher.

AHM was formed earlier this year as unease grew within the mathematics community over how fast frontier labs were moving into the field. That unease has only intensified since, with 25 Fields Medal winners, including Terence Tao, signing a declaration warning that AI labs racing to solve famous math problems for benchmarks is causing a “severe misalignment” with how the field actually makes progress. A separate Leiden Declaration, signed by thousands of researchers, has made similar arguments about AI companies failing to properly credit the human mathematicians whose published work trains their models.

AHM itself has been the most combative of these groups. Earlier this month, it called out OpenAI directly over the company’s release of more than 700 AI-generated mathematical manuscripts, saying mathematicians never asked for the work and that the release amounted to a demonstration of power rather than a genuine contribution to the field. The association urged mathematicians to discontinue their work with OpenAI altogether.

Not every prominent mathematician is on board, however. The field has also seen the opposite pull, with labs successfully recruiting some of its biggest names. Fields Medal winner Jacob Tsimerman joined OpenAI this year, predicting that AI will soon do everything mathematicians do “better and faster”, a view that sits in sharp contrast to AHM’s stance. Executives at the frontier labs themselves have made similarly bold predictions, with Google DeepMind’s David Silver previously saying he’d be amazed if AI mathematicians don’t end up transforming the whole of mathematics.

That split is essentially the fault line AHM’s growth reflects: a widening gap between mathematicians who see frontier AI labs as an existential threat to how the field defines progress and credits work, and those who are willing to work inside the labs to shape how AI handles mathematics going forward. With over 1,000 members and a public, weekly-updated list of who has pledged what, AHM is betting that the former group is only going to keep growing.