The 2006 Fields medalist — often called mathematics' top honor — and his field's most decorated living practitioner, is asking his colleagues to organize and reshape the discipline before AI finishes rewriting it.
Mathematics has built its reputation on centuries of patient, solitary proof. Terence Tao, the 2006 Fields medalist widely regarded as his field's most decorated living practitioner, told colleagues at this week's International Congress of Mathematicians in Philadelphia that the timeline for figuring out what AI means for the profession has just collapsed to months. His argument: the response has to be political, not technical.
Tao made the case in a New Scientist interview on the eve of the congress. The work of being a mathematician (what counts as a contribution, who gets credit, how peer review runs, which problems are worth pursuing) is being rewritten from the outside faster than the profession can rewrite it from within. He is asking his colleagues to stop waiting on that rewrite and to do it themselves, before the end of 2026.
"Mathematicians need to organize, become activists, get a little political," Tao told the magazine. The demand he is making is concrete: build institutions that decide how AI tools enter the discipline, who audits their outputs, and which kinds of proof remain the province of human judgment.
Throughout 2026, AI models have been solving decades-old open problems at what Tao characterized as "a few each week," a pace that, in a discipline where a single breakthrough can take a career, reads less like an upgrade than a regime change. Some of those solutions, Tao noted, are coming from amateurs and hobbyists prompting models directly. The bottleneck of training and credentials that once gated the field is starting to leak.
Tao calls this the first peril of comparable severity to hit mathematics in more than a century, a deliberate comparison to the early twentieth-century turn toward abstraction and rigor. The current shock changes who, or what, produces the theorems.
The "political" part is what makes the demand unusual coming from a field that prides itself on objectivity. Tao is not asking mathematicians to campaign for office. He is asking them to set standards for AI-assisted proofs, push journals and funders to require disclosure of how a result was derived, and treat the integration of AI tools as a governance problem rather than a productivity upgrade. The implicit argument is that if mathematicians do not build those structures themselves, the structures will be built for them by funders, publishers, or model vendors, and the discipline's standards of patience and proof will not survive the transition intact.
Every knowledge profession under automation pressure faces the same fork: write the new rules, or inherit someone else's. Tao's intervention is a public attempt to keep the first option on the table for mathematics. "Mathematicians need to reimagine what it means to be a mathematician," he said. The verb matters: reimagine, not defend.
The next concrete test is whether mathematics publishes an AI-proof disclosure standard before the year ends. Tao has put a clock on that decision at the congress in Philadelphia, and he is betting that a field built on patient, solitary proof can also build its own rules under deadline.