AI math claims raise questions about proof checking and credit

In short: AI systems are being credited with major math breakthroughs, but many of the most famous claims still need careful, independent checking.

What's going on

Several AI related math results reported in 2026 have sparked excitement and anxiety, including fears that mathematicians could be replaced. But mathematicians interviewed by The New York Times said the situation is more complicated.

In May 2026, OpenAI reportedly produced a counterexample to the Unit Distance Conjecture, a problem first raised in 1946. A counterexample is like finding one clear case that breaks a rule someone thought was always true. Researchers then used the key idea from that work to disprove another conjecture.

In September, OpenAI said an unreleased model had solved the Navier–Stokes Millennium Prize problem, one of seven famous problems with a $1 million prize for a correct solution. Mathematicians said the reported proof was long and dense, about 166 pages, and hard to evaluate. Some also questioned the exact scope of what was claimed and what assumptions were used.

In October, OpenAI released hundreds more claimed results, 722 manuscripts grouped into 372 “result families,” across areas like number theory and geometry. Experts cautioned that it could take months or years to sort out which results are correct, new, and important.

What to watch

A key point is that “solving” a math problem can mean different things. A result might be correct but still hard for humans to check, learn from, or build on, like getting an answer without showing clear work. Watch for independent verification and for new tools and norms about who gets credit when AI is part of the discovery process.

Source: NYTimes