New Funding to Help Mathematicians Verify AI Research
OpenAI and Anthropic join the Simons Foundation to provide researchers with tools to audit automated mathematical proofs.
Updated on Oct. 15, 2026 in Mathematics
Live Poll
Should private AI companies provide funding for independent academic review of their research outputs?
OpenAI and Anthropic have pledged a combined $15 million to a new Simons Foundation program designed to fund mathematicians assessing AI-generated research. OpenAI contributed $5 million to the effort, which will provide computing resources and AI tools to researchers through an open call for proposals.
Why it matters
As AI models increasingly contribute to mathematical research, establishing rigorous verification standards is essential to maintain accuracy. This funding initiative addresses the challenge of auditing machine-generated proofs and findings, a problem underscored by the recent withdrawal of three manuscripts due to a sign error.
The current math collection contains 719 total manuscripts across 372 result families. This follows the October 7, 2026, withdrawal of three manuscripts discovered to contain a sign error.
The players
OpenAI
An AI research and deployment company focused on large language models and automated reasoning systems.
Anthropic
An AI safety and research company that develops large-scale generative models and interpretability tools.
Simons Foundation
A private philanthropic organization that funds scientific research, specifically supporting mathematics and basic science.
Institute for Advanced Study
A research center for theoretical study and intellectual inquiry that currently hosts an AI and math advisory group.
The details
The program supports independent assessment of AI-generated work through workshops, conferences, and specialized research programs. By providing researchers with direct access to AI tools and computing resources, the initiative seeks to build a pipeline for verifying formal proofs. This methodology targets the verification of complex logic and automated outputs before they are accepted as established results.
Timeline
September 2026: OpenAI announced an independent mathematics and AI advisory group at the Institute for Advanced Study.
October 6, 2026: The math collection held 722 manuscripts across 372 result families.
October 7, 2026: Three manuscripts were withdrawn from the collection following the discovery of a sign error.
The Tech Race
This initiative follows the recent formation of an independent mathematics and AI advisory group at the Institute for Advanced Study. It marks a shift from purely theoretical AI advisory boards to a resource-heavy model that integrates active computing power into the verification process.
Professional mathematicians and researchers will soon have an avenue to request computing resources and AI tools via the Simons Foundation's upcoming open call for proposals. These resources are intended to help standardize the verification of automated findings within the mathematical community.
The takeaway
The integrity of machine-generated mathematics depends on the development of robust, independent verification workflows. Watch the Simons Foundation's upcoming open call for proposals to see which research teams receive access to these new AI tools and computing resources.
Further reading
For broader context on how research institutions are auditing new computational tools, see Mathematics.
Live Poll
Should private AI companies provide funding for independent academic review of their research outputs?
