Artificial intelligence is rapidly transforming the landscape of scientific research, with mathematics among the disciplines experiencing profound disruption. In early September 2026, OpenAI announced that an internal AI model had solved the Navier-Stokes existence and smoothness problem, one of seven Millennium Prize problems that had remained unsolved for decades. This marked the first time a machine had resolved a Millennium Prize problem and only the second such problem solved overall. The AI reportedly arrived at a solution within 88 hours.
Subsequently, on October 6, OpenAI released a comprehensive set of 719 papers detailing solutions to at least 372 other longstanding open problems, a development that prompted reactions across the mathematical community. While the scale of the achievement is unprecedented in any scientific field, many prominent mathematicians expressed deep concern. Hugo Duminil-Copin, a 2022 Fields Medal recipient, described the event as causing “immense shock” and paralysis within the discipline.
Duminil-Copin and 27 other Fields medallists issued a statement warning of a “severe misalignment of AI in mathematics.” They cautioned that solutions generated without human comprehension might pose risks to the future of the field. Martin Hairer of EPFL in Lausanne echoed these sentiments, suggesting that unchecked AI intervention could “sterilise” mathematics by sidelining the intellectual process that underpins discovery.
Opponents of relying solely on AI-generated solutions emphasize that the value of mathematics lies not just in answers but in the journey toward them, which often leads to the creation of innovative tools and concepts that have far-reaching applications. For instance, the concept of imaginary numbers, once considered impractical, later became fundamental to numerous scientific and engineering fields. Critics argue that the AI-generated papers lack the kind of novel insight or new research directions that propel the discipline forward. Luis Martínez-Zorado of CUNEF University, a specialist on the Navier-Stokes problem, described the AI’s 166-page explanation as a fragmented compilation rather than rigorous research.
The impact of AI extends beyond mathematics. A January 2026 study analyzing over 41 million scientific papers found accelerating adoption of AI across natural sciences. Theoretical physics, in particular, has embraced AI tools to expedite calculations, with researchers turning to large language models such as OpenAI’s and Anthropic’s Claude to advance their work. Notably, Anthropic reported progress in modeling complex particle interactions in late September.
While AI use correlates with increased publication rates and citations, the same study noted a reduction in the diversity of research topics by nearly 5% and a 22% decline in cross-disciplinary engagement. These trends raise concerns about a narrowing of scientific inquiry. A recent Google survey of more than 600 scientists found that nearly half felt AI encourages a focus on safer, incremental projects with predictable outcomes, while only a minority believed it fosters more speculative research, which is often critical for breakthrough discoveries.
Looking ahead, researchers like Hector Zenil of King’s College London foresee a future where AI systems might autonomously conduct the entire scientific process—from hypothesis generation to experimentation and evaluation—potentially producing results that elude human understanding. While such machine-driven science could offer valuable findings, it would represent a fundamental shift in how knowledge is pursued, with scientists potentially serving as interpreters of AI-generated outcomes rather than originators of new ideas.
Despite unease, mathematicians are not obsolete. The opaque nature of current AI solutions necessitates extensive verification and interpretation, ensuring a continuing role for human expertise. Moreover, the creative aspect of mathematical discovery remains elusive for AI, with uncertainties about whether machines could generate entirely novel concepts like imaginary numbers.
As Alain Connes, a leading mathematician who once doubted that computers could tackle Millennium Prize problems, acknowledges, the rapid advancement of AI challenges long-held assumptions. The coming months and years may reveal just how deeply artificial intelligence will transform science in ways both promising and unsettling.