Mathematicians now have AI dietary labels

Some mathematicians have decided the best way to handle AI is to leave it off the plate.

What happened. The Association for Human Mathematics (AHM) says it now has 1,008 members, 654 of whom have also joined its optional "AI-free caucus." Membership is free, public, and open to people in mathematical research at graduate level and above. New Scientist reported on October 20 that membership includes professors and PhD students.

The two pledges, plus one. Every member commits to two things: not working with commercial AI companies (technical labor, math expertise, consulting or publicity), and not publishing mathematical writing generated by AI, from papers and referee reports to lecture notes and books. Members of the AI-free caucus go further and don't use AI models in research at all: no prompting AI to find proofs, explore examples or pose questions. The AHM's FAQ says the commitments are "expressions of principle rather than precise legal agreements."

Vegetarians and vegans. New Scientist describes the two groups that result as "vegetarians" (the two required pledges) and "vegans" (who also skip AI in research). Yale's Sam Raskin told the magazine he's a vegetarian member who didn't sign the third pledge because it was slightly unclear whether his occasional use for simple tasks counted. Imperial College London's Steven Sivek, a vegan member, said "We're not denialists" and that the group is working on how the profession should adapt.

Why now. New Scientist links the group to a backlash over how AI companies have released math results, including OpenAI publishing 722 papers at once earlier this month. The AHM's own statement on that release says: "Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power."

Our take

The menu joke writes itself, but the complaint underneath is a real one about reviewing work, not a fear of computers. A proof that a machine has checked can still be impossible for a human to learn anything from, and a pile of 700 of them lands on the same volunteer reviewers. Whatever you think of the pledge, "show your work in a form people can read" is a fair rule for AI output in any field, including code.

What we did and didn't do: this story is based on New Scientist's report and the AHM's own members, FAQ and statements pages, read on October 20, 2026. We haven't independently checked the member count beyond the AHM's own list.

MODEL Editors