A growing fight between some of the world's leading mathematicians and artificial intelligence companies has brought an uncomfortable question to the centre of the AI debate: what happens when machines begin solving problems that humans have spent decades trying to understand? Twenty-five winners of the Fields Medal, widely regarded as the highest honour in mathematics, have warned that the race among AI companies to tackle famous unsolved problems could cause lasting damage to the discipline.The joint letter, published Friday on mathematician Terence Tao's blog, was signed by some of the biggest names in modern mathematics. The signatories include Tao as well as mathematicians from nearly five decades of Fields Medal history, from Pierre Deligne, who received the medal in 1978, to Yu Deng, who was honoured this year.Their concern is not that artificial intelligence should stay out of mathematics. It is about what mathematics is supposed to achieve in the first place.Why are mathematicians angry with OpenAI?The dispute intensified after OpenAI said earlier this week that an unreleased AI model had solved the Navier-Stokes problem, one of the seven Millennium Prize Problems. The company said its system worked on the challenge for 88 hours, with thousands of AI agents operating in parallel.For mathematicians, however, the question goes far beyond whether an AI system can produce a result that appears to solve a famous problem.The 25 Fields Medal winners said the objectives of AI companies and the mathematical community are currently "severely misaligned". They warned that the issue could become "a general threat to intellectual work". The mathematicians argue that solving a problem is often only part of the purpose of mathematical research.Problems are deliberately created and selected because they force students and researchers to develop ways of thinking, understand concepts and discover connections. The final answer is important, but the journey towards it can be just as valuable. "Solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight," the mathematicians wrote.Those processes, they added, "invariably take time and are based on human interaction." That distinction is at the heart of the current dispute.The Navier-Stokes controversyThe disagreement erupted after OpenAI announced its claimed breakthrough involving the Navier-Stokes equations. The equations are among the most important mathematical tools used to describe the movement of fluids and gases. A rigorous solution to the Millennium Prize version of the problem has remained out of reach for generations.OpenAI said its AI agents had worked on the problem in parallel and produced what the company described as a solution. But the timing immediately became controversial.Just hours before the announcement, NYU mathematician Tristan Buckmaster and Anthropic's Levent Alpoge had released their own AI-assisted work involving related equations. Buckmaster later accused OpenAI of moving onto the problem after learning about research he and Alpoge had been developing for months. He also alleged that the company had copied their unusual approach.OpenAI denied having access to their work. The dispute has added another layer to a much bigger argument: when an AI system appears to make a mathematical breakthrough, who is responsible for checking the result, explaining it and establishing whether it actually works? Human mathematicians, critics argue, cannot simply take an AI company's announcement at face value.What is "slop mathematics"?Some mathematicians have begun using the term "slop mathematics" to describe what they see as a growing problem. The criticism is that AI systems can generate large numbers of possible mathematical results at extraordinary speed, but human researchers may then have to spend significant time checking those claims.If a result is wrong, researchers have to identify the mistake. If it is promising, they may have to develop a rigorous proof. And if it is genuinely important, they may ultimately have to explain its significance to the wider mathematical community. The academics behind the criticism say that this work is often done without compensation, credit or acknowledgement."After the latest result is dropped, research mathematicians are compelled to step in to properly verify, disseminate, and sometimes discredit entirely the claimed results," the academics said. "This labor goes uncompensated, uncredited, and unacknowledged." That is one reason the debate has become increasingly heated.OpenAI pulls support for Caltech mathematics eventThe disagreement has also spilled into the academic world. OpenAI has withdrawn its support for a Caltech mathematics event after current and former mathematicians criticised the growing use of AI in mathematical research. The event, described as a "mathathon", was designed to give young mathematicians an opportunity to use AI tools while working on difficult, open mathematical problems.OpenAI was among the companies supporting the event. But the sponsorship came under criticism after Caltech mathematicians issued an open letter warning that the programme "is likely to have destructive impacts for the mathematical community." OpenAI research lead Dan Roberts said Thursday, September 10, that the company would no longer sponsor the event.How does the Caltech mathathon work?The programme was designed around the idea of using AI to augment, rather than replace, human mathematical research. According to its application information, participants first compete in a 40-hour round. They receive $20,000 worth of AI tokens and work on an open mathematical problem.Teams that produce answers considered promising and well explained can progress to a second stage. That phase lasts six months and provides participating teams with additional AI resources to continue their work. The stated goal is to "responsibly use AI tools to augment human understanding of mathematics."But the controversy shows how difficult that balance may be. For researchers worried about the direction of AI in mathematics, the concern is that the technology could gradually change what gets valued. Speed, novelty and impressive demonstrations may become more important than patient understanding, collaboration and proof.OpenAI says it wants to engage with mathematiciansOpenAI has acknowledged that its rapid progress in mathematics is creating disruption. Roberts said the company recognised that the development was disruptive and that OpenAI wanted to work more closely with mathematicians. "We recognize that the rapid progress of AI in mathematics is disruptive," Roberts said. "We're looking to engage with the math community more on the best way to integrate this technology and communicate its impacts."That response could become increasingly important as AI systems move deeper into scientific research. The issue is no longer simply whether an AI can solve a difficult equation. It is about how the result is presented, who verifies it, who receives credit and whether the process helps humans understand mathematics better or simply produces answers faster.The bigger question: What should AI do for mathematics?The Fields Medal winners ended their letter with a question that extends well beyond mathematics. As AI changes how intellectual work is carried out, how can society make sure it does not lose sight of what that work was originally intended to achieve?For mathematicians, that means preserving the human elements of the discipline: understanding, insight, discussion, teaching and the slow process of discovering why something is true. AI may eventually become an extraordinarily powerful mathematical research partner. There is little doubt that it can search possibilities and perform calculations at a scale humans cannot match.But the current argument suggests that the mathematics community does not want the future of the discipline to be defined simply by whoever reaches an answer first. The battle, in other words, may not really be mathematicians versus AI. It may be a fight over what counts as progress in an age when machines can increasingly produce answers before humans have had time to understand the questions.