AI-Assisted Mathematician Claims Breakthrough Proof of Riemann Hypothesis

Levent Alpöge, a mathematician at AI company Anthropic, asserts he has proven the Riemann Hypothesis using an LLM named Claude, submitting a proof coded in the Lean verification language. The hypothesis, proposed in 1859 by Bernhard Riemann, concerns the distribution of prime numbers and has remained unsolved for 167 years. If verified, the proof would mark a pivotal moment in mathematics, signaling AI's transformation from a tool to a superhuman collaborator in the field.
On November 20, 2026, Levent Alpöge, a mathematician at AI company Anthropic, announced he had achieved a proof of the Riemann Hypothesis—a conjecture first proposed by Bernhard Riemann in 1859—using an AI system named Claude. The proof, submitted in the formal verification language Lean, is now under scrutiny by the mathematical community. The Riemann Hypothesis, which posits that all non-trivial zeros of the Riemann zeta function lie on the critical line with real part 1/2, has remained unproven for 167 years and is considered the most significant open problem in number theory. Alpöge’s claim marks a pivotal moment in the evolution of AI in mathematics, where AI has transitioned from a supportive tool to a potential co-author of a landmark proof. The use of Lean, a language designed to verify mathematical correctness, adds a layer of rigor, though the verification process by human experts is still underway. Alpöge’s statement reflects both personal reflection and a broader philosophical shift, acknowledging the emotional weight of potentially closing a centuries-old quest while expressing awe at the new possibilities AI enables. The outcome remains uncertain: while the proof’s validity is not yet confirmed, its existence challenges long-held assumptions about human limits in mathematical discovery.
If verified, Alpöge’s claim could redefine the role of AI in scientific advancement, shifting perceptions from automation to collaborative intelligence. Mathematicians may face a recalibration of their professional identity, as AI systems begin to contribute to foundational discoveries once thought uniquely human. Institutions like universities and research labs may reevaluate how they train and assess researchers, potentially integrating AI collaboration into curricula and peer review. The broader scientific community may see a ripple effect, with other fields—such as physics or biology—reassessing their reliance on human intuition versus machine-assisted insight. A comparable precedent is the 1994 proof of the Four Color Theorem using computer assistance, which initially sparked controversy but eventually led to wider acceptance of computational methods. However, unlike that case, this claim involves not just computation but a deep, creative reasoning process attributed to an LLM. What remains to be seen is whether the mathematical community will accept AI-generated proofs as valid without full human comprehension, and how institutions will adapt to a future where AI contributes to the most profound breakthroughs. The next steps will likely involve peer review, formal verification, and a broader debate on the nature of mathematical understanding.