Generative artificial intelligence (AI), the type used in tools like ChatGPT and Claude, has shown impressive abilities in mathematics. Advanced models from companies like OpenAI and Anthropic have demonstrated the capacity to solve complex mathematical problems in recent months. However, Terence Tao, a renowned mathematician and 2006 Fields Medalist, has raised concerns about the role of AI in solving mathematical problems. In a public discussion on Mastodon on September 3, Tao suggested that allowing AI to solve mathematical problems on its own may not always be beneficial for the field. He warns that open mathematical problems—those that have not yet been solved—could become "non-renewable resources" due to what he calls "contamination." Once an AI solves a problem, the clarity and value of the problem itself may be permanently reduced because the process by which the AI reached the solution is often unclear. Tao’s concerns echo those of another Fields Medalist, Hugo Duminil-Copin, who warned in late August that using AI to automatically solve open problems could harm the development of new mathematical ideas and techniques. Duminil-Copin compared this to using heavy machinery to dig up artifacts from an archaeological site, destroying the historical and contextual value of the findings. In a similar vein, Tao cited a case study from group theory, a branch of mathematics, where he and others had worked to measure the difficulty of a problem by preventing participants from sharing solutions. Once key information was shared to help progress, they felt the original "mapping" of the problem’s difficulty had been irreversibly damaged. Tao acknowledges that enforcing restrictions on AI use in mathematics would be extremely challenging, given the widespread availability of such tools and the incentives for quick progress in research. However, he believes that social pressure could still discourage some overreliance on AI in problem-solving. He even suggests that certain categories of math problems should be off-limits to automated solvers to preserve their value for training future mathematicians and advancing the field as a whole. These concerns extend beyond mathematics, raising broader questions about the role of AI in areas like art and other professions. While no one suggests banning AI entirely, the idea is to protect certain spaces for human creativity and development. This is not Tao’s first warning about AI—earlier, he has criticized the use of AI-generated proofs in mathematics, highlighting the challenges they pose to the integrity and understanding of mathematical research.