Simons Foundation bets $15M that mathematicians — not the AI labs — should write the rules for AI in math
The Simons Foundation is teaming up with Anthropic and OpenAI on a $15M effort to help mathematicians evaluate AI-assisted results and set community standards — insisting the rules come from the community, not the labs.
The announcement
The Simons Foundation announced on Thursday a new set of initiatives aimed at helping the mathematics community navigate AI’s rapid incursion into the field. Under a partnership with Anthropic and OpenAI, the three will commit $15 million in combined funding and resources. The foundation will manage and award the funding — including through an open call for proposals — to support 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.
The money matters less than the message attached to it. “The community should be driving discussions of how these tools are evaluated, adopted and used, not any single company or institution,” said foundation president David Spergel. “The goal of the program is to support the mathematics community.”
What the money funds
The program is deliberately community-driven: open calls, workshops, and initiatives rather than a single flagship grant. The stated priorities are evaluating AI-assisted results — a polite way of saying check the proofs the machines produced — training researchers on new tools, and building shared standards for when and how AI can be used in mathematical work.
There is a practical reason to fund this now. AI systems are generating ever more plausible-looking mathematical arguments, and mathematics is the one discipline where a wrong answer can look exactly like a right one until somebody verifies the chain. A community that can’t evaluate machine-assisted proofs doesn’t have a trust problem — it has a standards problem. Fifteen million dollars aimed at standards is money spent at the right joint.
The compute clause
Anthropic and OpenAI aren’t just contributing funding and goodwill: both labs will provide AI tools and computing resources to researchers supported by the program — and that support extends beyond mathematics into basic science. The labs get structured feedback from some of the world’s most rigorous evaluators; the mathematicians get compute budgets and frontier-model access that grants in pure mathematics almost never cover.
The Lean backstory
The Simons Foundation isn’t arriving at this intersection late. It was an early backer of machine learning for science — including funding the development of the Lean proof assistant, the interactive theorem prover that has become the standard substrate for machine-checkable mathematics. In a sense, this program closes a loop: the foundation helped build the tooling that made AI-assisted proofs legible, and now it is funding the community norms for using them.
Why it matters
Mathematics has spent two millennia as the one domain with an unambiguous standard of correctness: a proof is either right or it isn’t. AI is about to stress-test that standard at scale, generating far more candidate arguments than any human community can referee. Somebody has to decide what counts as an acceptable machine-assisted proof, how it gets checked, and who gets a say. The Simons Foundation’s bet is that the answer should be the mathematicians themselves — with the labs writing checks and supplying compute, not the rules. Spergel’s other line from the announcement reads like the program’s thesis: “Now is the time to look at mathematics as a whole — not as a collection of proofs and theorems, but as a collective understanding driven by human curiosity.”
Sources: The Simons Foundation’s official announcement (Oct 15, 2026); EIN Presswire distribution of the announcement (Oct 15, 2026).