The relationship between OpenAI and parts of the mathematics community is becoming increasingly tense as AI companies compete to solve some of the field's most difficult and famous problems.

Twenty-five prominent mathematicians have now signed an open letter warning that the rapid development of AI-powered mathematical systems could threaten the way mathematical discoveries are created, credited, verified, and shared. Every mathematician who signed the letter has received the Fields Medal, widely regarded as the highest distinction in mathematics.

Mathematicians Raise Concerns About AI Research

The concerns come as leading AI laboratories increasingly use powerful language models to tackle advanced mathematical problems.

This week, NYU professor Tristan Buckmaster accused OpenAI of putting pressure on him not to properly recognize a collaborator affiliated with Anthropic who had contributed to solving an important mathematical problem.

Buckmaster also questioned whether OpenAI could have used work produced through Codex to help develop its own recently announced proof of a major mathematical problem during an intensive period of AI inference.

OpenAI Faces Criticism at Caltech

The disagreement has also affected OpenAI's relationship with academic institutions.

On Thursday, the company withdrew its sponsorship of a mathematics event at Caltech after researchers at the university criticized OpenAI's approach.

The developments illustrate how quickly tensions are growing between AI companies and researchers who are concerned about the effects of increasingly capable AI systems on established academic practices.

The Problem With AI-Generated Proofs

AI systems capable of solving previously unsolved mathematical problems could potentially provide enormous benefits to science and humanity.

However, the mathematicians behind the open letter argue that simply producing a solution is not enough.

Mathematical discoveries need to be carefully explained, independently examined, connected to existing research, and communicated to other mathematicians before they can become part of the broader body of mathematical knowledge.

The signatories argue that AI-generated solutions are sometimes announced so quickly that there is little opportunity to properly document the methods used, identify genuinely new ideas, or acknowledge previous contributions.

Attribution and Plagiarism Concerns

The situation also creates difficult questions surrounding intellectual ownership.

If an AI system produces a breakthrough based partly on research created by human mathematicians, determining who deserves recognition can become complicated.

The mathematicians argue that attribution is especially important in creative disciplines such as mathematics. Without proper credit and documentation, researchers could see their contributions absorbed into AI-generated discoveries without receiving appropriate recognition.

They also emphasize that human mathematicians play an essential role in developing, interpreting, teaching, and integrating new mathematical ideas into the existing body of knowledge.

Growing Fear Around Codex and AI Training

The controversy has also caused some researchers to become concerned about how their interactions with AI coding and research tools might be used.

Some mathematicians are reportedly wondering whether work they produce while using Codex or similar AI systems could eventually contribute to the development of future models.

This uncertainty could have a serious impact on the culture of open academic research.

Researchers traditionally share ideas and preliminary findings because collaboration and open discussion help scientific knowledge advance. If mathematicians begin worrying that sharing their work with AI tools could allow technology companies to reproduce or commercialize their discoveries, they may become more reluctant to share their research.

AI Could Change the Incentives for Discovery

The financial resources available to major AI laboratories could further change the competitive landscape.

If an AI company identifies a promising approach to an unsolved mathematical problem, it can potentially spend millions of dollars on computing resources and run large-scale AI systems until they produce a solution.

That creates a new incentive: instead of researchers publicly collaborating on difficult problems, organizations may have reasons to keep promising discoveries secret until they can produce and announce a solution first.

Such competition could fundamentally change how mathematical research is conducted.

The Leiden Declaration

The latest open letter follows the Leiden Declaration, which was published by a group of mathematicians earlier this year.

That document examined how large language models and AI-generated proofs could affect mathematics and proposed recommendations for researchers, academic institutions, and policymakers.

The broader discussion is part of a growing effort to determine how the mathematical community should adapt as AI becomes capable of performing increasingly sophisticated research tasks.

Mathematics Is More Than Just a Proof

The debate highlights an important distinction between producing a mathematical answer and advancing mathematics as a discipline.

A proof is only one part of the process. Mathematics also depends on developing new concepts, identifying important questions, teaching future researchers, connecting separate areas of knowledge, and building a shared intellectual framework.

Human mathematicians play a major role in determining which discoveries matter and how those discoveries influence future research.

The concern is that an AI system could potentially generate technically correct results without providing the broader intellectual context that allows those results to become meaningful to the mathematical community.

A Warning for Other Professions

The concerns raised by mathematicians extend well beyond mathematics.

Software engineers, scientists, writers, designers, researchers, and other professionals are all experiencing similar changes as AI systems become integrated into their workflows.

The central question is no longer simply whether AI can perform a particular task. It is also about how society can ensure that technological progress does not undermine the principles, institutions, and human contributions that give that work its value.

The mathematicians behind the letter argue that the challenges appearing in their field could provide an early warning for other professions.

As AI continues to reshape how knowledge and creative work are produced, the larger challenge may be finding a balance between faster discovery and preserving the human systems of collaboration, attribution, understanding, and knowledge-sharing that make those discoveries valuable.