Mathematician Protest Reaches 11,000 Signatures Across Three Declarations as Community Pivots to Intellectual Response
The mathematician response to AI’s encroachment on research mathematics has scaled past 11,000 signatures across three declarations. The Leiden Declaration, which launched in early June and crossed 2,300 signatories within weeks, now has 4,000+. The Math and AI declaration (mathandai.org) has 7,000+. A separate letter from Fellows of the Royal Society focused on AI existential risk landed on September 16.
OpenAI’s announcement of a solution to the Millennium Prize variant of Navier-Stokes existence and smoothness, and the subsequent credit dispute involving an Anthropic researcher, triggered what one participant called “a host of reasoned declarations and open letters, spiking in intensity.” The Leiden Declaration had already crossed 3,000 signatories by late July before the Navier-Stokes announcement in early September; the OpenAI claim accelerated a movement that was already growing.
From Protest to Framework
Three months in, the movement is producing more sophisticated intellectual responses alongside the petitions. On September 19, Terence Tao hosted a guest essay by Po-Shen Loh, Carnegie Mellon professor and former US Mathematical Olympiad coach, that argues the declarations are using the wrong framework.
Loh’s case: the existing declarations are organized around protecting incumbency. They treat the question as “how do mathematicians preserve their role?” That framing is both strategically weak and philosophically insufficient. Economists Tyler Cowen and Joshua Gans each challenged it: Gans concluded it was “a loss of control from incumbents in a scientific field,” while Cowen rejected “the most cynical interpretations” but still very much differed with the declarations’ approach.
His alternative rests on a single axiom: “We (humans) should help humanity flourish.” Applied to mathematics, the reframe shifts the question from “should AI replace mathematicians?” to “what is the human-optimal configuration of AI-assisted mathematical communities?” The second question has a defensible answer independent of whether any particular job category survives.
Loh further argues that AI-driven productivity will generate such a volume of necessary human roles, in verification, interpretation, education, and domain integration, that labor demand will outpace supply. His conclusion: mathematicians do not need to fight AI adoption because the economics of AI-driven human specialization will preserve human mathematical communities as a byproduct.
What the Declarations Actually Say
The three main declarations differ in emphasis:
Leiden Declaration (4,000+ signatories): focuses on the epistemic integrity of mathematical proof, specifically the risks of AI-generated proofs that cannot be fully verified by human reviewers. The concern is epistemic, not economic.
Math and AI declaration (7,000+ signatories, hosted at mathandai.org): titled “A Severe Misalignment of AI in Mathematics,” it covers AI’s effects on mathematical training, publication norms, and credit attribution. The Navier-Stokes authorship dispute sits directly in this territory.
Royal Society Fellows letter (September 16): positions the concern within the larger frame of AI existential risk. The signatories emphasize their concern is for everyone, not only mathematicians, a rhetorical move that directly addresses the incumbent-protection critique.
The 25 Fields Medalists who signed their open letter on September 11 accused AI labs of “severe misalignment with mathematics.” That letter prompted the Cowen and Gans responses that Loh’s essay engages.
Signatories Growth
The Leiden Declaration’s trajectory reflects a movement that had broad support before the Navier-Stokes announcement. It crossed 2,300 signatories by late June and 3,000 by late July, suggesting the mathematical community’s underlying concern had been building across the summer independent of any single event. The September announcement added intensity but did not create the movement from scratch.
Whether the declarations translate into any change in lab behavior or funding policy is an open question. What the mathematical community has is credibility and the ability to shape how AI-generated mathematical claims are received by the broader scientific and funding community, which is arguably what the Leiden Declaration was designed to influence.