The Advisory Group on Mathematics and AI (Sep 29, 2026) recommends how frontier labs should release AI-generated math results: deposit promptly, cite related work, formalize where possible, disclose prompts and costs, and fund community-led human understanding.
Responsible Release of AI-Generated Mathematics
September 29, 2026 — Advisory Group on Mathematics and Artificial Intelligence
At present, some frontier AI labs are testing advanced mathematical problems on proprietary models that remain inaccessible to the broader scientific community. Our recommendations are formulated with this practical context in mind. However, ideally, they would not do so. We want to state clearly from the start: we do not endorse this practice, and we ask them to stop testing advanced mathematical problems on proprietary models.
1. Background
The mathematical community has long-standing norms concerning the dissemination and peer review of results. One of the most important scholarly norms in mathematics is that the authors of a paper should understand the mathematical argument of that paper, have verified its correctness themselves, and take full responsibility for the content. Moreover, authors of works that significantly advance the field regularly give seminars and conference talks, explaining their new developments.
However, it is now the case that AI can output mathematical arguments in situations without the human who prompted it being able to understand the arguments, verify them, or take responsibility for them. We believe that human understanding of mathematics remains of paramount importance.
We asked the mathematical community for feedback about what it would mean for AI labs to responsibly release mathematical results and received over 600 replies. Informed by these responses, we arrived at a set of recommendations aimed at any AI lab whose models are likely to have a significant impact on mathematics. Overarching principles:
If AI labs produce significant mathematical results, they should responsibly release the results as soon as possible.
AI labs that release substantial mathematical output without immediate accompanying human understanding must take responsibility for ensuring that human understanding will follow—including funding.
The development of human understanding must remain organic and community led. It should not be directed by AI labs.
2. Responsible release of results generated by AI labs
2.A. Papers that a human understands
Papers for which there is a mathematician responsible who fully understands the content should follow traditional norms: post a preprint, submit for peer review, and give talks.
2.B. Papers that are not yet understood by anybody
Step I: Initial release
Labs (not later mathematicians) should (a) scour the literature for related ideas and cite them, and (b) produce a write-up in the style of a traditional mathematical paper with friendly introductions and precise theorem statements.
Deposit results in timely scholarly repositories not controlled by any AI lab, with persistent identifiers and version history. Do not treat releases primarily as marketing.
For each result, make public the model name, prompts used, a summarized chain of thought, time taken, and estimated compute cost.
Formalize proofs where possible; clearly state formalization status when delayed.
Document how AI came to be used on each problem, and how many comparable problems the models tried and failed.
Step II: Supporting human understanding
Labs should provide support, including funding, for community-led activities such as conferences, workshops, working groups, postdocs, and expository writing. Acceptance of such support does not confer legitimacy on proprietary testing practices.
3. Ensuring broad access
Proprietary internal models risk a two-tier system that alienates mathematicians from their own discipline. Unequal access to publicly available models risks exacerbating institutional inequalities. Labs should grant the global mathematical community broad, equitable access to their publicly available models.