Build1 publisherNot yet confirmed elsewhere3 min readPublished
Mathematicians' group chaired by Terence Tao calls for an OpenAI boycott over 700 AI proof files
Mathematicians in a group chaired by Terence Tao are urging a boycott of OpenAI after it released more than 700 AI-generated proof files at once. At issue is whether a problem counts as solved before a human has checked and understood its proof, and who does that work.
The Engineer · Build desk
What happened
- According to The Decoder, many of the papers are so dense and oddly written that leading experts need AI help to decipher them, if they can at all.
- The AHM says OpenAI ignored the central advice of AGMAI, its advisory group at the Institute for Advanced Study, that advanced problems should not be tested on internal models.
- OpenAI had earlier claimed an internal model solved more than 100 open math problems in one month, including an approach to the Navier-Stokes Millennium Problem.
- OpenAI's Navier-Stokes announcement came shortly before two mathematicians who had used ChatGPT presented their own solution, and OpenAI denied training on those chats.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- cost Anyone who wants to build on one of the released proofs first has to make it readable, and OpenAI's release model leaves that unpaid work with the mathematical community.
- constraint Because AGMAI has no say over the pace of OpenAI's internal research, withholding cooperation is the leverage left to mathematicians who object to how results are released.
- constraint Sharing a promising partial idea now risks someone with a model finishing it first; Tao says researchers are already keeping such directions to themselves.
Two questions are being argued under the word "solved" [24]. The first is whether every step of a proof is correct. The second is whether a human can explain why it works and answer questions about it. A machine-checked formal proof would settle the first and leave the second open. The Decoder's report does not say whether any of the more than 700 files include one [2].
Tao's side holds that the second question decides it [10]. He previously joined 24 other Fields Medalists, 25 in all, in a statement warning of a "severe misalignment" between the AI industry's goals and those of mathematics [9][25]. Problem-solving, they argued, is only a tool for the discipline's real goal of conceptual understanding [10]. Mass-producing solved problems at an ever-faster pace could "destroy fertile ground instead of breathing life into new ideas," they wrote [11]. They also said rushed announcements without proper write-ups or citations of prior work raise "severe attribution and plagiarism questions" [12].
On Mastodon, Tao described how that understanding used to spread. A proof of a long-standing open problem led to talks, workshops, new collaborations and textbooks, and young researchers came in to work through it [13]. Now, he says, problems are being solved autonomously by AI users who have no interest in the field. Those users cannot understand the results well enough to answer questions or give talks, and far fewer seminars and collaborations follow [14]. His "Math 2.0" is an era in which solving problems is no longer the discipline's main focus [15].
The AHM statement applies that view to the batch release. It calls the release of more than 700 files at once "not a demonstration of scholarship, but a demonstration of power" [5]. "Mathematicians did not ask for this work to be done," the statement reads [6]. Anyone who maintains an open source project and has been sent an unsolicited pull request of that size will recognise the complaint. The statement's opening paragraph also raises OpenAI's copyright lawsuits, implying without saying so that the results depended on training on mathematicians' work [7].
Scott Aaronson, who called the situation a "Mathocalypse" on his blog [8], set OpenAI's approach against what he calls the "Anthropic model" [21]. Anthropic teamed up with two algorithm researchers, and the AI model in that collaboration supplied the key idea behind disproving two conjectures that were each decades old. Both were paid and wrote a version of the proof that humans could follow [21]. For a field whose stated goal is understanding, I think that is the better default, because the party that produced the result paid for the readable version [10][21]. The Decoder points out its cost: a private company chooses which mathematicians act as "emissaries" for a result [23].
What to watch
- Whether any of the 700-plus files are independently checked and written up in human-readable form, and who ends up doing that work.
- Whether OpenAI answers the boycott call or changes how it releases results, including through AGMAI.
- Whether more mathematicians or institutions join the AHM in refusing to work with OpenAI.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+30
- Incentives65
- Confidence40
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
The Association for Human Mathematics (AHM) is calling for a boycott of OpenAI, accusing it of violating core norms of scientific research and urging mathematicians to stop working with the company.
- [2]
OpenAI published hundreds of AI-generated mathematical manuscripts, releasing more than 700 files at once.
- [3]
Many of the papers are so dense and oddly written that even leading experts need AI help to decipher them, if they can at all.
- [4]
Fields Medalist Terence Tao chairs the AHM and shared its statement as a guest post on his blog.
- [5]
The AHM statement describes the release of more than 700 files at once as "not a demonstration of scholarship, but a demonstration of power."
- [6]
"Mathematicians did not ask for this work to be done," the AHM statement reads.
- [7]
The AHM statement's opening paragraph brings up OpenAI's copyright lawsuits, suggesting without saying so directly that the company could only achieve these results because it trained on mathematicians' work.
- [8]
Complexity theorist Scott Aaronson described the situation as a "Mathocalypse" on his blog.
- [9]
Tao previously joined 24 other Fields Medalists in signing a statement warning of a "severe misalignment" between the AI industry's goals and those of mathematics.
- [10]
The Fields Medalist signatories argue that problem-solving is only a tool for the discipline's real goal of conceptual understanding.
- [11]
The signatories wrote that mass-producing solved problems at an ever-faster pace could "destroy fertile ground instead of breathing life into new ideas."
- [12]
The signatories criticize rushed announcements of AI-generated solutions without proper write-ups or citations of relevant prior work, saying this raises "severe attribution and plagiarism questions."
- [13]
Tao wrote on Mastodon that in traditional mathematics, proofs of long-standing open problems led to talks, workshops, new collaborations and inclusion in textbooks, attracting young researchers who worked to understand and contextualize them.
- [14]
Tao says problems are now being solved autonomously by AI users who have no interest in the field and cannot understand the results well enough to answer questions or give talks, producing far fewer seminars and collaborations than traditional breakthroughs.
- [15]
Tao outlined a "Math 2.0" era on Mastodon in which solving problems should no longer be the main focus of the discipline.
- [16]
Tao says researchers are holding back promising directions for fear that competitors will scoop the solution.
- [17]
OpenAI earlier claimed an internal model solved more than 100 open math problems in a single month, including an approach to the Navier-Stokes Millennium Problem.
- [18]
In response to criticism, OpenAI set up the advisory group AGMAI at the Institute for Advanced Study; the group explicitly has no say over the pace of the company's internal research.
- [19]
The AHM says OpenAI ignored the central premise of AGMAI's advice, which held that advanced math problems should not be tested on internal models.
- [20]
OpenAI announced a solution to the Navier-Stokes problem shortly before two mathematicians could present their own AI-assisted solution; the two had used ChatGPT, and OpenAI denied suspicions that it used data from those interactions to train its system.
- [21]
Aaronson contrasts OpenAI's one-batch release with what he calls the "Anthropic model": Anthropic partnered with two algorithm researchers whose AI model supplied the key idea for disproving two decades-old conjectures, and the researchers were compensated and wrote a version of the proof humans could follow.
- [22]
OpenAI's method leaves the mathematical community doing the thankless work of making proofs readable, for free.
- [23]
Anthropic's method lets a private company choose which mathematicians get to serve as "emissaries" for a result.
- [24]
A dispute is taking shape over whether a problem can count as solved before its proof has been independently worked through and made understandable to humans.
- [25]
The earlier misalignment statement was signed by 25 Fields Medalists in total, counting Tao.
Sources
1 independent publisher whose own reporting we read for this story.
- the-decoder.comSome mathematicians call for OpenAI boycott after AI-generated proofs flood their field
2 articles · October 8, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Research attribution and verificationFollow
- AI in mathematicsFollow
Entities
- OpenAIFollow
- Advisory Group on Mathematics and Artificial IntelligenceFollow
- Association for Human MathematicsFollow
- Institute for Advanced StudyFollow
- Unique games conjectureFollow
- Scott AaronsonFollow
- Terence TaoFollow
- AnthropicFollow
- Navier-Stokes existence and smoothness problemFollow