OpenAI has published 722 mathematical papers produced largely by an unreleased artificial intelligence model, covering 372 open problems across algebra, geometry and theoretical computer science, The Conversation reported on 9 October.
The release — variously labeled a "drop," a "dump," a "carpet bombing" and a "mathocalypse" by researchers — has unsettled the mathematical community. Several papers claim significant advances on high-profile problems, including the Riemann hypothesis and the Birch–Swinnerton-Dyer conjecture, each carrying a US$1 million bounty as one of the seven Millennium Prize Problems.
The episode amounts to a stress test for the centuries-old system by which mathematics establishes truth: expert human review. If AI can generate plausible-looking but flawed proofs at scale, the bottleneck in the discipline may shift from discovery to verification.
Retractions, amendments and a "serious hallucination"
Cracks appeared quickly. OpenAI has already retracted three of the papers due to an elementary error and amended several others after mistakes invalidated their results, according to The Conversation.
A senior mathematician who found one of his favourite problems among those claimed solved told The Conversation the accompanying paper was so unintelligible that, had he received it as an editor at a mathematics journal, "it would have gone straight into the bin."
Even OpenAI's own model, GPT-5.6 Sol, was skeptical when asked about the trove, describing at least one high-profile result as "a serious hallucination … [which] should never be cited, submitted, or circulated as a proof without a complete expert audit," The Conversation reported.
OpenAI said its motivation for the release was to "enable further progress in mathematics." Critics counter that the manner of the release — hundreds of papers at once, bypassing journals entirely — works against that goal, since the traditional refereeing of long, technical papers can take years.
Did AI solve the Riemann hypothesis?
Not yet, by any accepted standard. The papers claim advances on the Riemann hypothesis and the Birch–Swinnerton-Dyer conjecture, but no claimed solution has survived expert audit, and the retractions suggest insufficient vetting before publication.
The Millennium Prize Problems are seven long-standing challenges, each with a US$1 million bounty for a verified solution. Verification, not generation, is the hurdle — and that remains human work.
Can Lean formalization verify AI-generated proofs?
OpenAI claims to have formalized 300 of the main results using software systems such as Lean, which build mathematics from basic axioms so that each logical step can be machine-checked. The approach, sometimes called autoformalization, has gained traction over the past decade; an early milestone was the 2008 formalization of the four color theorem.
But formalization is no guarantee. Past evidence, including issues with the Navier-Stokes formalization, suggests AI-generated formalizations can themselves contain errors. Whether the mathematical community will accept OpenAI's Lean formalizations as sufficient verification remains an open question.
A September controversy and a US$1.4 trillion backdrop
The release follows a disputed episode in September, when OpenAI published a claimed solution to a case of the Navier-Stokes problem, another Millennium Prize challenge. Mathematicians Tristan Buckmaster and Levent Alpöge, who had been working on the problem themselves, alleged OpenAI accessed their data and used their ideas, then devoted some US$15 million of computing power to find a solution. OpenAI has denied the allegations.
Observers also link the trove to OpenAI's pursuit of artificial general intelligence, or AGI — a hypothetical AI surpassing humans in virtually all cognitive tasks. Research-level mathematics is widely seen as evidence of progress toward that goal, and such claims could benefit the company ahead of a planned stock market launch reportedly targeting a valuation of up to US$1.4 trillion.
Community reaction has split sharply. Some researchers called the release "obviously the most significant moment in mathematical history," while one unnamed researcher described it as "a giant pile of turd dumped at our doorstep."
Specialists are now combing through the trove. How many more papers are retracted or amended, whether any Millennium Prize claims survive audit, and whether the Navier-Stokes allegations are pursued further will determine whether the "mathocalypse" yields lasting advances or a cautionary tale about automated reasoning.