Twenty-five winners of the Fields Medal, the highest honor in mathematics, published a joint statement warning that the goals of the AI industry and those of mathematics are "severely misaligned." The signatories include Terence Tao, Pierre Deligne (1978), Peter Scholze (2018), Maryna Viazovska (2022), Martin Hairer (2014), Cedric Villani (2010), Manjul Bhargava (2014) and this year's winner Yu Deng. The reason this matters beyond one discipline: the authors describe a pattern that, in their words, all of humanity might face as AI changes how work is done.
What the statement says
The signatories do not dispute the technical progress. Large language models have become good enough at mathematics in recent months to crack "major outstanding problems in many fields of mathematics." That capability is precisely what worries them. In their view, AI companies treat mathematical problems as benchmarks to conquer, and that approach damages both the science and the community around it. The statement lands amid a controversy between two mathematicians and OpenAI: the researchers accuse the company of hearing rumors about a partial solution to a Millennium Prize Problem and trying to beat them to the result for publicity. OpenAI chief researcher Pachocki had said during the Astra announcement that the company deliberately chose not to optimize the model for math; shortly after, OpenAI apparently trained math models anyway, seemingly in direct response to those rumors.
The core of the argument is about what a solved problem is worth. Famous unsolved problems, the statement says, "have often served as landmarks and lighthouses against which one can measure an improved understanding of this landscape." When someone cracks one, the answer matters less than the thinking required to get there. Mathematicians then spend years taking that thinking apart in what the signatories call a "long and arduous process of talks, discussions, simplifications." AI threatens to short-circuit exactly this stage: "Solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight," they write. Flooding the field with answers at machine speed could, in their wording, "destroy fertile ground instead of breathing life into new ideas."
The mechanics of that damage are already visible in how AI results enter the field. Solutions get announced with "no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others," which raises "severe attribution and plagiarism questions." The statement also points to the human chain that turns a result into accepted knowledge: "Without the willing mathematicians who must take care of their development and integration into the mathematical canon, AI-conceived ideas would never become fully alive," it reads, and "the crucial human transmission chain between mathematicians would be lost." The signatories frame this as a "general threat to intellectual work": years of training have "served not only to produce a final answer or product, but also to develop understanding and the ability to formulate new questions and ideas," and when AI produces "the results of such work directly," those purposes come apart.
What this means for business
For companies adopting AI, the statement is a reminder that a benchmark score and a usable capability are different things. The mathematicians are not calling for a ban: AI "offers the potential of enhancing and accelerating genuine mathematical study and understanding," and the profession will have to adapt. But "whether these changes ultimately benefit the field or have a destructive effect will in large part be determined by the decisions of the humans in control of this new technology." The same logic applies to any team that replaces a junior workflow with a model: the output arrives, while the ability to judge it, extend it and explain it does not necessarily follow.
The gap is easiest to see in education, where homework can increasingly be done by AI while exams still ban it, and the distance between those two performance levels, measured in grades, keeps growing. A research paper from the NATO Special Operations University recently described a related pattern as the "tragedy of the cognitive commons": each company that replaces entry-level jobs with AI reaps efficiency gains, but the cost of eroding expertise is spread across the entire talent pool. What the researcher describes across whole professions, the mathematicians are already watching play out in their own field. For a small company the trade-off is sharper, because it has less slack to rebuild lost expertise; for a large one the risk is slower and more diffuse, accumulating across teams that no longer train replacements.
What the statement does not settle is measurement. It offers no threshold for when AI assistance stops building understanding and starts substituting for it, and no mechanism for attributing machine-generated results. Those questions are what a vendor conversation should cover: how the tool records and cites prior work, who on the team can still verify an answer without it, and what happens to the reasoning chain when the model is unavailable. The signatories say "these issues must be addressed urgently" and call on the mathematical community, the companies building these tools, and "a society that will confront similar problems in many other forms of intellectual work."
The marker to watch is the signatory list itself. It opened with 25 names and already includes winners from 1978 through 2026; if it keeps growing across disciplines and the companies named in the attribution dispute respond with concrete documentation practices rather than statements, the concern has moved from an internal debate to a procurement criterion. Until then, the practical reading for a business is narrower: treat every AI-generated result as a draft whose value depends on whether anyone in the room can still reconstruct the reasoning behind it.
