OpenAI says it is contributing $5 million to a new Simons Foundation initiative that helps mathematicians examine results produced by AI. The foundation announced the effort on Oct. 15. With Anthropic and OpenAI it commits $15 million in combined funding and resources.
What the money pays for
The foundation manages and awards the funding, including through an open call for proposals. Its page says the money supports workshops and initiatives that help researchers "evaluate recent AI-assisted results, learn and use new tools, share perspectives and shape collective standards for AI use."
OpenAI's update banner names "workshops, conferences, and special programs." Anthropic and OpenAI will also give foundation-supported researchers AI tools and computing resources, and the foundation says that reaches beyond mathematics into basic science.
The foundation's president, David Spergel, says "the community should be driving discussions of how these tools are evaluated, adopted and used, not any single company or institution."
Where the $5 million figure comes from
The amount appears only in OpenAI's banner, added to its Oct. 6 post on Oct. 15. The Simons page gives no split between the companies and does not state Anthropic's share. It describes the $15 million as combined "funding and resources," so the page does not say how much of OpenAI's part is cash.
What mathematicians objected to
The banner is attached to the post in which OpenAI released new results from an internal frontier model. That post already said OpenAI would fund workshops, conferences and special programs around understanding them. The repository's README now lists 719 manuscripts in 372 families, and about 42% of top-line results formalized. It warns that some unformalized results could have issues.
The Association for Human Mathematics responded the next day. Its statement, reposted on Terence Tao's blog, says "mathematicians did not ask for this work to be done." It calls releasing over 700 files at once "a demonstration of power" and urges mathematicians to stop working with OpenAI.
The statement's mention of "accusations of illegal plagiarism" describes lawsuits OpenAI is defending. It does not accuse OpenAI of plagiarizing the Oct. 6 results.
A separate dispute, still contested
A different fight started in September. When OpenAI announced a Navier-Stokes proof on Sept. 8, NYU professor Tristan Buckmaster alleged, TechCrunch reported, that information about his progress with Levent Alpöge, an Anthropic employee, had been passed to OpenAI. It is his claim, and we have not read his own statement.
OpenAI denies using their work. Its current page says its effort began Sept. 1 after a rumor it later tied to the two, that it saw none of their work before they published, and that an investigation found Buckmaster's Codex prompts could not have influenced its system. The denial was revised Sept. 10.
Does the fund answer the criticism?
Our reading: it answers part of it. Control over awards sits with the foundation rather than OpenAI, both pages say so, and the stated purpose includes evaluating AI-assisted results and shaping standards. In our view that speaks to a worry that a lab would grade its own work.
It does not touch the objection to how the results were released, or the association's call to stop working with OpenAI. Also our reading: money from the companies whose output is examined will draw scrutiny whatever the safeguards, and the Simons page says nothing about award criteria, timing or Anthropic's share.
What we could not check
We did not read OpenAI's X post about the funding, or statements from Buckmaster or Alpöge, or the association's own site, only the Tao repost of its text. Claims about those come from the pages linked above.
