OpenAI publicly released 722 internal mathematical manuscripts generated by AI models—only to withdraw three and revise 14 the very next day after a sign error came to light. To tackle the problem that human verification and comprehension capacity cannot keep pace once AI produces mathematical results in bulk, the Simons Foundation, together with OpenAI and Anthropic, has jointly invested funds and resources worth a total of $15 million to support mathematicians in verifying, understanding, and extending mathematical results produced by AI—with funding led by the mathematics community. The field of mathematics is currently moving from proof scarcity to proof abundance. Long derivations rapidly generated by AI are difficult for humans to read and understand quickly, and academia is divided over AI companies dumping large quantities of insufficiently verified results. This collaboration responds to the academic community's earlier demand that AI laboratories provide supporting resources to help humans digest AI results, and the related support will subsequently be extended to basic scientific research fields beyond mathematics.