AI Solved a Millennium Problem — and the Math World Is Panicking
Twenty-five mathematicians, including a Fields Medal winner, have issued an emergency statement against AI solving hard math problems — calling it harmful to science and the mathematical community. OpenAI's claim to have cracked a Millennium Prize Problem has exposed a deep fault line in academia.
A Statement That Sounds Frantic
On September 11, twenty-five mathematicians around the world — including Shigefumi Mori, a Fields Medalist and special professor at Kyoto University — released what can only be described as an emergency statement condemning artificial intelligence for solving difficult mathematical problems.
The statement calls the practice “harmful to science and the mathematical community.” It was issued in direct response to an OpenAI announcement on September 8 claiming that its developing AI system had solved one of the seven Millennium Prize Problems — mathematical questions so profound that each carries a $1 million reward and solving any of them would represent a watershed moment in human intellectual history.
What makes this episode striking is not the mathematics involved but the visible panic on the part of established researchers. These are people who spent their lives mastering the very craft that an AI company now claims it can perform faster than humans ever could.
The Statement: A Defense of What Math Actually Is
The wording of the declaration is telling. The mathematicians describe mathematics as “a microcosm of humanity” — a phrase that immediately elevates the discipline beyond calculation into something almost spiritual. They argue that the goals of AI companies are “significantly divorced” from the true purpose of mathematical research.
The core concern, stated plainly, is that solving problems conceptually is not the same as solving them computationally. “Problem-solving is merely a means to achieve the original purpose of conceptual understanding and insight,” the statement reads. The distinction is one between doing mathematics and understanding mathematics — and it is a distinction that feels increasingly irrelevant in an era where AI systems can generate valid proofs that human experts struggle to verify.
The statement raises several specific fears. First, if AI produces “results” in mass quantities, human verification will fall behind. Second, the environment that allows time-intensive thinking — the kind of sustained contemplation that characterizes genuine mathematical breakthroughs — risks being lost. Third, because AI derives answers from existing vast bodies of work, it “creates problems related to attribution of achievements and plagiarism.”
Each of these concerns has weight. But they also reveal something uncomfortable: the mathematical establishment is facing the same existential uncertainty that other knowledge professions have confronted in recent years, and it is reacting with declarations rather than adaptation.
Who OpenAI Actually Solved
OpenAI claimed its system solved one of the Millennium Prize Problems. These problems were posed by the Clay Mathematics Institute in 2000 and include questions such as the Riemann Hypothesis and P versus NP. They have resisted solution for over two decades despite the efforts of some of the brightest minds in the field.
The claim, if verified, would represent a genuine milestone. But verification is precisely the problem the mathematicians raise. How does a human expert confirm that an AI-generated proof is correct when the proof itself may exceed human cognitive capacity? The statement suggests the answer is: you can’t, not quickly enough, and that delay creates a dangerous gap between what is claimed and what is known.
This tension between computational power and human verification is not unique to mathematics. It appears in drug discovery, climate modeling, and legal analysis. But mathematics is special because mathematical truth is absolute — a proof is either correct or it is not. There is no room for the probabilistic approximations that other fields accept. When AI generates a proof, the stakes of error are uniquely high.
The Real Story: A Fracture Over What Counts as Discovery
Western media will likely treat this incident as a footnote — a brief exchange between an AI company and a group of irate academics. But the deeper significance is a fracture running through the mathematical community itself over what counts as real discovery versus mechanical computation.
For centuries, mathematics has been understood as a human endeavor. The beauty of a proof lies not just in its conclusion but in the path taken to reach it — the intuitions, the failed attempts, the moments of insight that connect one idea to another. This is what the Fields Medal celebrates: not just results but the human capacity to see what others cannot see.
AI challenges this conception at its foundation. If a machine can generate a valid proof without understanding it in the human sense, does the proof still count? Is the result less true because it came from a different kind of intelligence? These are not abstract questions. They are questions that will shape funding, careers, and the direction of research for decades to come.
The mathematicians’ statement is ultimately a plea for the protection of something that cannot easily be quantified: the human dimension of mathematical discovery. Whether that plea will carry weight against the momentum of commercial AI development is an open question.
What Happens Next
The statement has drawn attention primarily within academic circles, but its implications extend far beyond. Universities that fund mathematical research will need to decide whether to accept AI-generated proofs, whether to credit AI systems alongside human researchers, and how to train the next generation of mathematicians in a world where computational tools are increasingly powerful.
OpenAI and other AI companies will face pressure to be more transparent about their methods and to collaborate with the mathematical community rather than announce results unilaterally. The current pattern — announce first, verify later — is unsustainable when the claims involve truths that must stand the test of rigorous scrutiny.
Most importantly, the mathematical community itself must decide whether it will adapt its practices to incorporate new tools or resist them entirely. The statement suggests the latter instinct dominates among senior researchers, but resistance alone cannot preserve the discipline. The question is not whether AI will change mathematics but how.
The thirty-eight hours it took OpenAI to produce its claimed result, according to some reports, is already a number that will be cited in debates for years. Whether it represents the beginning of a new era in mathematical research or a cautionary tale about overreach remains to be seen. What is clear is that the conversation has begun, and the mathematicians who signed the statement have made their position unmistakably plain.