A proof produced by artificial intelligence and announced earlier this month as part of a batch of 377 AI-derived mathematical results is winning over the human experts who have spent the past two weeks taking it apart. Writing on October 19, Quanta Magazine reported that number theorists have now unpacked the AI-generated proof of the quasi-Riemann hypothesis — a weaker version of the Riemann hypothesis, arguably the biggest open question in mathematics — and their verdict is unusually emphatic.
Jared Duker Lichtman of Stanford University called it "the most significant AI-for-math achievement so far, by quite a margin." Maksym Radziwiłł of New York University, whose own work underpins parts of the proof, described the implications as "staggering." For more context on this story, see our ongoing latest AI developments.
What the Quasi-Riemann Hypothesis Claims
The Riemann hypothesis has been open for more than 160 years. It concerns the Riemann zeta function and the locations of its nontrivial zeros, and it is intimately connected to how prime numbers are distributed among the integers. Because primes are the building blocks of arithmetic, a proof would reshape huge areas of number theory.
The hypothesis is one of the seven Millennium Prize Problems named by the Clay Mathematics Institute, and it was the only problem carried over from David Hilbert's famous list of 23 problems announced a century earlier.
The technical skeleton, as Quanta lays out, has been mapped step by step for generations. Riemann showed in 1859 that the zeta function's nontrivial zeros govern how primes thin out along the number line, and proved they can only sit inside a vertical strip where the real part of the input lies between zero and one. In 1896, two mathematicians independently showed that no zeros lie on the strip's right edge — a result strong enough to prove the prime number theorem. The hypothesis itself says all nontrivial zeros sit exactly on the strip's midline, where the real part equals one half. Trivial zeros at negative even integers are already fully understood. More than 160 years of effort has narrowed and probed that strip without ever pinning the zeros down.
The quasi-Riemann hypothesis, or quasi-RH, is a strictly weaker statement. Even so, its consequences reach nearly as far. "In a sense, the vast majority of consequences of RH already follow from quasi-RH," Radziwiłł told Quanta.
From Announcement to Verification
The result first surfaced on October 6, when OpenAI announced 377 mathematical results derived with its AI systems. Among professional mathematicians, the announcement initially met with measured skepticism — AI systems have previously produced claimed proofs that collapsed under expert scrutiny, and verification has long been the bottleneck separating AI-generated mathematics from accepted mathematics.
According to Quanta's report, this proof is different: unlike other recent AI proofs, this one has been unpacked by human experts. Number theorists including Radziwiłł, Alexander Dunn of the Georgia Institute of Technology, and others have been working through the argument since the announcement, checking its steps and situating it within existing knowledge.
The checks have held up. "The proof is a surprising synthesis of well-known ideas," said Dunn, who along with Radziwiłł pioneered techniques on which the result relies in part.
Old Ideas, Hidden Connections
Perhaps the most striking aspect of the proof, mathematicians told Quanta, is what it did not require. Contrary to expectations that settling even a weakened form of the Riemann hypothesis would demand fundamentally new machinery, the AI-produced proof introduced no new concepts. Instead, it unearthed hidden connections between established ones, combining techniques in a way no human had assembled.
That pattern — novelty through unexpected synthesis rather than through new theory — is emerging as a signature of AI-assisted mathematical discovery, and it is prompting researchers to reconsider where machine contribution to mathematics is most likely to arrive first.
Why It Matters for AI Research
The verified result is a milestone for AI for mathematics, a field that has accelerated rapidly as reasoning models have improved. OpenAI's October 6 announcement, which we covered when the Institute for Advanced Study raised questions about the batch of results, was the largest single claim of AI-derived mathematics to date. The quasi-RH proof is the first result from that batch to receive a detailed expert read-through, and it has survived.
For AI labs, the episode demonstrates a workflow that may become standard: models produce candidate proofs at scale, and human experts triage, verify, and absorb the results that matter. For mathematicians, the immediate consequence is narrower but real — a body of consequences of the Riemann hypothesis that no longer need the full conjecture to hold.
Quanta notes that mathematicians are only beginning to work through what the proof means for the field. As Radziwiłł put it, the implications are "staggering" — an adjective rarely attached to a result that stops short of the full hypothesis.
The episode also sets a benchmark for future claims. Any AI lab announcing mathematical breakthroughs will now be measured against quasi-RH: a claimed result that was checked, verified, and accepted by the very experts best positioned to refute it.
None of this settles the Riemann hypothesis itself. The full conjecture remains open, and quasi-RH is strictly weaker. But for the first time, an AI system has produced a proof at the frontier of analytic number theory that leading specialists in the area have unpacked and endorsed — a fact that may say as much about the future of mathematical research as the theorem itself.
This article is based on reporting by Quanta Magazine, published October 19, 2026.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
