On September 21, 2026, OpenAI announced on X that it is collaborating with an independent mathematician advisory group to 'responsibly share progress in AI and mathematics.' According to the official statement, the advisory group will advise on three things: how to evaluate and communicate new mathematical results, how to uphold academic and professional standards, and how to build tools that support mathematical research and learning. OpenAI also said it wants mathematicians at the center of 'shaping how AI supports mathematical understanding and how its benefits reach the broader community.'
These three duties actually point to the same class of problem: when an AI system produces a mathematical result, who judges whether it holds, by what standard it is publicly released, and how it is accepted by the mathematical community after release. OpenAI's wording places 'evaluation' and 'communication' side by side, indicating its concern is not only whether the result itself is correct, but also how the result enters public view.
What needs to be distinguished is that all of the above are statements from OpenAI's official post and are institutional self-description; whether the advisory group is actually operating, how frequently it intervenes, and whether it has veto power over specific results are all unaddressed in the available information.