AnswersPublished21 sources
Why the AHM wants mathematicians to cut ties with OpenAI
The Association for Human Mathematics says OpenAI’s release of hundreds of mathematical manuscripts is not a substitute for scientific review and has urged mathematicians to stop working with the company. OpenAI published proof materials and Lean formalizations for some results, but not every manuscript has been for...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- The Association for Human Mathematics says OpenAI’s release of hundreds of mathematical manuscripts is not a substitute for scientific review and has urged mathematicians to stop working with the company.
- OpenAI published proof materials and Lean formalizations for some results, but not every manuscript has been formalized or independently verified.
- The controversy is about potential mathematical progress as well as transparency, the effort required to check the results, and the place of human researchers in the process.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: Why has the Association for Human Mathematics (AHM) urged mathematicians to stop working with OpenAI after its October 6 release of AI-gener. Article summary: AHM urged mathematicians to stop working with OpenAI because it sees the October 6 release as a mass display of a company’s capability, not a responsible way to establish mathematical knowledge. It argues that hundreds o. Topic tags: general, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, clic
The Association for Human Mathematics (AHM) has urged mathematicians to stop working with OpenAI following the company’s October 6, 2026, release of a large collection of mathematical manuscripts. Its objection is not simply that some proofs may be wrong. The group argues that a bulk release from a corporate AI model cannot replace the scrutiny, documentation and discussion that make research credible.45
The collection may contain valuable results. But a large number of papers, supporting proof files and headline-grabbing claims do not amount to independent confirmation. The dispute also asks who can examine the model, how its outputs can be reproduced, and who has to do the work of checking them.
What OpenAI released
OpenAI’s initial announcement described 722 manuscripts across 17 areas of mathematics, grouped into 372 families of related results. That means the collection is not 722 separate, independent discoveries. The public repository includes manuscripts and supporting proof artifacts and is released under the Apache-2.0 license.7
15
OpenAI said the model was tested on about 4,000 problems and that the average computing time per result was equivalent to roughly three hours of ChatGPT Pro Thinking. The company also published 10 shortened summaries of the model’s reasoning and Lean formalizations for some proofs.7
12
16 Lean is a system that can let a computer check formalized proofs. The repository says the results are at different stages of verification and that not all have Lean formalizations. Its current catalogue lists 719 manuscripts, rather than the 722 in the initial announcement.
10
11
The collection includes claims related to the Mahler conjectures and the exact irrationality exponent of π. Those claims should not be treated as conclusively established simply because they appear in published manuscripts.5
21
Why the AHM objects
In its statement, the AHM said mathematicians “did not ask for this work to be done” and described the release of hundreds of files as a “demonstration of power,” not scholarship. It called for skepticism and for science centered on people, rather than a rush to accept or process a company’s outputs.45
There is also a practical problem of scale. Specialists need to read the manuscripts, check the arguments and compare the claims with existing research. Making the files public gives the community material to examine, but it does not make that work disappear. Much of the burden falls on researchers outside the company.
The AHM also pointed to a warning from the Advisory Group on Mathematics and Artificial Intelligence, an independent group established to advise AI companies on their engagement with mathematical research. According to the AHM’s statement, the group had cautioned against testing advanced mathematical problems on companies’ internal models.45
What Lean checks—and what it doesn’t
Lean formalizations can provide a powerful check: they allow a computer to verify proofs that have been translated into the system. OpenAI says the main result of 162 manuscripts has been formalized.5 That adds support for those specific results, but it does not automatically verify every claim in every manuscript.
For work without a corresponding formalization, evaluation depends on people reading and checking the argument. The repository itself says the results are at different stages of verification and that not all come with Lean formalizations.10
11 So the collection should be assessed result by result—not treated as either entirely proven or entirely worthless.
More process details, but not full transparency
OpenAI’s figures on the problems attempted and computing time, together with its 10 reasoning summaries, offer some insight into how the work was produced.7
12
16 But they do not provide a full account of how each result was reached, and the model remains internal and unavailable for independent examination.
7
11
The total number of problems attempted also does not, by itself, show how often the model succeeded on a clearly defined set of difficult, open problems. To assess performance and reproducibility, researchers need more detail about the methods and the production of individual results.
OpenAI said it plans to hold workshops and conferences to examine important results.7 Such efforts could help researchers scrutinize the work, but they are not a substitute for publishing enough information for others to independently assess how the proofs were produced.
What the dispute means for mathematical research
There is no simple verdict yet. AI-generated arguments may contain genuine mathematical progress, but their value will depend on which results withstand independent scrutiny—not on the size of the release or the significance of the problems named.7
10
The AHM’s response also highlights questions about credit for earlier human work, the time required to evaluate new results, and whether attention or support could shift away from human-led research. The release does not resolve those concerns; it makes them harder to ignore.45
The central distinction is between an AI system’s ability to generate many mathematical claims and the process of establishing that those claims are correct. This debate is about both what AI might contribute to mathematics and what standards are needed to make that contribution checkable, transparent and meaningful to the research community.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Why the AHM wants mathematicians to cut ties with OpenAI"?
The Association for Human Mathematics says OpenAI’s release of hundreds of mathematical manuscripts is not a substitute for scientific review and has urged mathematicians to stop working with the company.
What are the key points to validate first?
The Association for Human Mathematics says OpenAI’s release of hundreds of mathematical manuscripts is not a substitute for scientific review and has urged mathematicians to stop working with the company. OpenAI published proof materials and Lean formalizations for some results, but not every manuscript has been formalized or independently verified.
What should I do next in practice?
The controversy is about potential mathematical progress as well as transparency, the effort required to check the results, and the place of human researchers in the process.