In a development that has sent ripples of both awe and apprehension through the global mathematical community, OpenAI convened a gathering of approximately 40 leading mathematicians in August to confront a profound question: what role remains for human intellect when artificial intelligence begins to eclipse human capabilities in the very realm of logic and proof? Sources present at the closed-door meeting revealed that the artificial intelligence research giant hinted at having achieved breakthroughs that could resolve hundreds of long-standing mathematical problems. However, company representatives assured attendees that these solutions would not be released en masse, a statement that OpenAI spokesperson Lindsay McCallum later clarified the company was "not aware of." The primary objective of the convocation, it seems, was to solicit advice from the very community potentially most impacted by these advancements, particularly concerning the ethical and logistical considerations of publishing such groundbreaking, AI-generated discoveries.
A Tense Dialogue: Excitement Meets Apprehension
The response from the assembled mathematicians was, by all accounts, a complex admixture of intellectual exhilaration and a deep-seated sense of unease. Bryna Kra, a distinguished mathematician from Northwestern University who attended the session, described the atmosphere as "a mixture of excitement and dread." She emphasized that the mathematicians urged OpenAI to adopt a more rigorous and traditional approach to dissemination than the company had employed for its earlier announcement of 10 AI-generated mathematical advances earlier that month. Instead of a mere blog post or a social media update, the consensus was that OpenAI should publish comprehensive research papers. This would allow the broader mathematical community the necessary time and space to meticulously absorb, critically analyze, and subsequently build upon these potentially transformative results. "Apparently, that input was ignored," Kra stated, reflecting a palpable sense of disappointment.
The Impending Floodgate: Hundreds of Solutions on the Horizon
Contrary to the initial assurances, OpenAI has reportedly informed individuals familiar with its plans that it intends to release hundreds of solutions to previously unsolved mathematical problems on GitHub this coming Tuesday. This impending deluge of AI-generated insights represents the latest in a surge of tens of thousands of mathematical solutions produced by artificial intelligence systems throughout the year, a testament to the escalating power and sophistication of frontier AI models.
"On August 28, we began training a new internal model," stated Lindsay McCallum, an OpenAI spokesperson. "In addition to resolving the Navier–Stokes Millennium Prize problem, this model has now resolved more than 100 long-standing open problems across most areas of mathematics. We are working to responsibly release the next math results from our model, drawing on advice and public recommendations from the Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study to inform how we release these results. We have not set a release time."
Echoes of Controversy: The Navier-Stokes Saga and Ethical Red Flags
For a significant contingent of prominent mathematicians, this impending release is viewed not as a sign of progress, but as a stark indicator that OpenAI may have failed to learn from the controversies that have marred its previous scientific disclosures. A palpable sense of frustration permeates conversations with leading academics, who perceive the current landscape as an arena where AI powerhouses like OpenAI and its rival Anthropic are engaged in a high-stakes competition to showcase their model’s prowess, particularly as both entities reportedly gear up for significant initial public offerings. In their haste to outmaneuver each other, mathematicians contend, the established protocols for scientific discovery, publication, and attribution have been systematically disregarded.
The most contentious incident to date occurred in September. OpenAI, after purportedly hearing "rumors" that other researchers were nearing a solution, deployed thousands of its AI agents to tackle a legendary, million-dollar Millennium Prize problem. This move drew sharp criticism from Tristan Buckmaster, a mathematician and professor at New York University. Buckmaster accused OpenAI of effectively "front-running" the work he had been conducting in collaboration with Levent Alpagut, an employee at Anthropic. While Buckmaster and Alpagut had not yet formally published their findings, they had been leveraging OpenAI’s own tools to aid their research.
Accusations of Coercion and Undue Influence
During an attempt to negotiate proper credit for their work, Buckmaster claims that OpenAI researcher Sébastien Bubeck appeared to suggest that Alpagut should be excluded from any co-authorship, citing that it would "make things complicated." In a statement to WIRED, Bubeck denied this assertion, pointing to a prior public declaration where he refuted any attempt to exclude Alpagut. Buckmaster, however, maintained that he refused Bubeck’s alleged proposition. Meeting notes reviewed by WIRED documented Bubeck expressing concern about Anthropic’s actions, reportedly stating, "There is some worry about what Anthropic is doing. Levent must be talking to his leadership right now letting them know fucking OpenAI can get a Millennium Prize problem. What’s to stop Anthropic from giving all their compute to get a Millennium Prize problem?"
Furthermore, according to the same meeting notes, Bubeck allegedly implied that Buckmaster might jeopardize his career if he were to publicly disclose his belief that OpenAI had appropriated his research. "If you don’t want me to be nice, then I don’t have to be nice," Bubeck is quoted as saying in the notes. Nestor Guillen, a visiting mathematics professor at NYU, articulated the sentiment felt by many in the academic community, describing the behavior of these AI companies as akin to "mobster behavior." OpenAI’s spokesperson, McCallum, vehemently disagreed with this characterization.
"I feel a lot of angst, and I see this more and more in my colleagues in mathematics, not over AI, but over the AI companies," Guillen remarked. "I think a lot of the angst is about the accumulation of power in one place."
The Advisory Group: A Step Forward or a Stalling Tactic?
In response to the growing unease, OpenAI established an advisory group in mid-September. The stated purpose of this group was to assist the company in refining its methods for assessing and communicating new discoveries, ensuring adherence to academic and professional standards, and developing tools that could genuinely support mathematical research and learning. However, mathematicians express that tangible progress from this initiative appears to be minimal.
Dissemination Dilemmas: Tweets vs. Peer-Reviewed Papers
A recurring source of frustration is the continued reliance by OpenAI and Anthropic on blog posts and social media for the dissemination of their findings, rather than the traditional and more robust scientific paper format. This approach not only hinders independent verification but also often fails to acknowledge or integrate prior contributions from other mathematicians. This trend extends beyond large corporations, with researchers like Alpagut himself announcing significant mathematical breakthroughs, such as disproving an 87-year-old conjecture, via a tweet shortly after a major sporting event.
"Math by tweet and math by press release to me is not the way to nurture the ecosystem that created the fertile ground that they have trained on," Kra asserted, highlighting the detrimental impact on the very environment that fostered AI’s development.
Community-Led Initiatives: Building Bridges in the Age of AI
In recognition of the escalating volume of machine-assisted proofs, several new tools and platforms have emerged to help the mathematical community navigate this evolving landscape. These initiatives aim to facilitate the adoption of new AI-generated results and to systematically categorize and verify mathematical discoveries. Notable among these are Hexagon, a repository dedicated to AI-generated mathematical content, and Palomar, a registry for machine-verified mathematics.
Despite direct encouragement from the community to integrate these resources, "they haven’t changed their behavior," Kra observed. She further expressed concern that OpenAI’s actions do not align with the principles outlined in the Leiden Declaration, a manifesto endorsed by over 4,000 mathematicians calling for AI companies to adhere to established academic standards—a declaration that Kra herself helped author.
The Specter of Obsolescence: A Divided Outlook
Compounding the ethical and procedural concerns is a disturbing undercurrent within OpenAI itself. According to individuals who have spoken with OpenAI employees, a segment of the company believes that its advanced technology has rendered traditional mathematics largely obsolete. "We don’t believe the future of mathematics is set," countered McCallum. "We’re working with the math community to navigate the future collaboratively." While the scenario of AI surpassing human research capabilities was presented as a hypothetical during the August meeting, attendees interviewed by WIRED interpreted it as a somber premonition of what is to come. The briefings provided by OpenAI to mathematicians about its capabilities and breakthroughs have been likened by one observer to the way "police might want to notify family before reporting a death in a car accident."
Sébastien Bubeck, however, appears to hold a more optimistic, albeit potentially controversial, perspective. A person familiar with his thinking suggests that Bubeck believes OpenAI’s technology will rapidly diminish the careers of most mathematicians. Bubeck himself expressed to WIRED his conviction that more advanced AI could empower mathematicians to tackle more ambitious problems, foster stronger connections between their work and real-world challenges, and ultimately enhance the accessibility of mathematics. "I see this as an opportunity to expand what mathematicians can do and the impact their work can have," he stated.
Mathematicians acknowledge the imperative to adapt, particularly in preparing future generations for this paradigm shift. However, they fundamentally reject the notion that their field is nearing its end. "It changes how we’re going to operate, but I think it’s a moment that we can think bigger," said Kra. She expressed a willingness to embrace powerful new tools for problem-solving, but stressed the critical need for greater transparency and rigor from AI companies to foster trust and enable the building upon their findings.
"It’s a scary time," Kra concluded, "but it’s also really a deeply exciting time."
Update: 10/6/2026, 2:22 pm ET: WIRED updated the story to clarify a reference to a set of math problems OpenAI has told mathematicians it has solved.
