AI's Riemann hypothesis proof is almost certainly true, mathematicians say
Independent mathematicians have all but signed off on an AI's claimed solution to math's most famous open problem, with final verification still running.

Independent mathematicians have attested that an AI-generated claimed proof of the Riemann hypothesis, the most important unsolved problem in mathematics, is almost certainly true, and researchers say their verification is near completion. For decision-makers, the episode sets the bar every AI vendor will now be measured against: a frontier claim that survives adversarial human review, not a benchmark score.
Independent mathematicians have attested that an AI's claimed proof of the Riemann hypothesis is almost certainly true, and the researchers checking the work say their verification is near completion. If it holds, the most important unsolved problem in mathematics, a question that has resisted every human attack since 1859, has effectively fallen to a machine. Sit with the second half of that sentence, because it is the part with teeth. Mathematics is the rare arena where a claim is either right or wrong and where the checking is designed to be merciless. A machine clearing that bar is not a parlor trick or a leaderboard score. It is a different kind of claim about what these systems can do.
For context, the Riemann hypothesis is a statement about prime numbers, the indivisible building blocks of arithmetic, and about how regularly they appear as you count higher. Bernhard Riemann, working in 1859, tied that question to a complex function called the zeta function and predicted that its zeros sit on a single vertical line. Nobody has proved it. It is one of the seven Millennium Prize Problems, the set the Clay Mathematics Institute flagged in 2000 with a $1 million award attached to each, and it is widely described as the most consequential of the group because so much of number theory is built on the assumption that it is true. A long list of results in the field begin with the phrase "assuming the Riemann hypothesis." Prove it, and a huge portion of mathematics stops being conditional.
The word doing the heavy lifting in the attestation is "almost." Independent mathematicians have signed off on the substance of the argument, but the formal verification is not finished, which is the normal rhythm of mathematics at this altitude. A proof is not accepted because it is elegant or because a lab says so. It is accepted when enough qualified people have read every line, checked every limit and failed to find the crack. That process is close to complete, not complete, and the gap between those two states is exactly where reputations get made and destroyed. The claim is that the argument is almost certainly true. That is an extraordinarily strong statement from people with nothing to gain from flattery, and it is still not the same thing as a finished, settled theorem.
What makes this a business story and not merely a beautiful one is the nature of the task. Large language models are excellent at retrieving and recombining what already exists. Producing a genuinely new proof of a problem that has been open for more than a century and a half is a different order of difficulty, because there is nothing to retrieve. The answer was not sitting in the training data waiting to be summarized. If independent mathematicians are right that this argument holds, then the evidence for machine reasoning stops being a demo and becomes a result that can be checked by people who are professionally hostile to hype. Mathematicians are the toughest audience on earth for a claim like this. They do not grade on a curve, and they have no commercial reason to be polite.
The second-order effects are where this gets interesting. A proof of the Riemann hypothesis is not, by itself, a skeleton key to the encryption guarding your messages, and nobody credible is suggesting otherwise. But the hypothesis sits underneath the deepest known structure in number theory, and number theory is the soil that modern cryptography grows in. More broadly, an AI system that can generate a valid new proof in one of the hardest domains humans have is an AI system that can plausibly generate valid new arguments in drug discovery, materials science, chip design and formal software verification, where being provably correct is worth real money. The value is not the theorem. The value is the demonstration that machine-generated reasoning can survive adversarial human review at the frontier.
Caveats deserve equal billing. Mathematics has a long history of claimed proofs of the Riemann hypothesis that did not survive scrutiny, and the phrase "almost certainly" is deliberately short of certainty. Verification is near completion, not finished. Until it is, the correct posture is disciplined curiosity rather than a triumphant press release. There is also an asymmetry that should worry anyone planning around this news: if the proof holds, the people who said so early look prescient, and if a flaw surfaces in the final pass, the people who repeated the claim too loudly own it permanently. The distance between attested and proved is where most bold claims go to die.
For executives, the practical lesson is about verification, not mathematics. Every industry is now being sold AI systems that produce fluent, confident, unverifiable output. The story worth tracking is the one where the output gets checked line by line by independent experts who have every incentive to find the error, and where those checkers come back and say it holds. That is the standard boards should demand when a vendor claims a breakthrough: who outside your payroll has verified this, and how close are they to done? Labs that can point to that kind of scrutiny will separate from the ones that can only point to a benchmark. Talent, capital and enterprise trust tend to follow whoever can prove it.
So watch the final verification pass. It is the difference between the most consequential mathematical result in a century and another cautionary tale about believing a headline too early. Either way, the fact that a machine is the one under review is the part that does not get reversed.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Science
Anthropic's I.P.O. Made Rank-and-File Researchers Worth Tens of Millions
Researchers who joined just a few years ago are now worth tens of millions, and the windfall resets the pay benchmark for every private A.I. lab still paying in paper.
Sea Lion Culling Strategy Fails: New Model Says Killing Them Won't Save Salmon
The latest research from AAAS overturn a long-held conservation assumption, forcing policymakers to rethink critical resource allocation in marine ecosystems.
Microplastics do more than pile up: they react with water to damage cells
Scientists have found that plastic particles can generate highly reactive molecules in water, a mechanism that could explain links to heart attacks, strokes, and neurological disorders.


