Human-AI Team Claims Breakthrough Proof of Generalized Riemann Hypothesis

American mathematician Levent Alpöge, working with Anthropic's AI Claude, has claimed to have proven the generalized Riemann hypothesis, a long-standing mathematical conjecture about prime number distribution. The proof, written in the Lean programming language for formal verification, has received preliminary validation from other mathematicians. While the result is a landmark in pure mathematics, its practical implications for technology, security, or daily life remain uncertain.
On November 20, 2026, American mathematician Levent Alpöge, a researcher at Anthropic, announced on his personal website that he and the AI system Claude had jointly achieved a proof of the generalized Riemann hypothesis (GRH), a landmark conjecture in number theory first proposed in 1859. The proof, formally written in the Lean programming language—a tool designed for machine-checked mathematical reasoning—has undergone preliminary validation by independent mathematicians, who have reported no immediate logical flaws. The GRH, a broader extension of the original Riemann hypothesis, concerns the distribution of non-trivial zeros of Dirichlet L-functions and has long been considered foundational to many areas of mathematics. Its resolution is one of the seven Millennium Prize Problems, with a $1 million reward offered by the Clay Mathematics Institute. Alpöge’s work is notable not only for its mathematical significance but for its method: the proof emerged from a collaboration between a human expert and an advanced AI, suggesting that AI may now be capable of generating original, rigorously verified mathematical knowledge. The broader context includes years of incremental progress in AI-assisted theorem proving, with earlier milestones such as DeepMind’s work on the Kashaev conjecture and the use of AI in formalizing proofs in systems like Coq and Lean. The involvement of Anthropic, a leading AI safety and alignment research lab, underscores the growing role of private tech firms in foundational science. What remains unresolved is whether the proof will withstand full peer review, whether the AI’s role was truly creative or merely synthetic, and what the long-term implications are for the future of mathematical discovery and human-AI collaboration in science.
If verified, Alpöge and Claude’s proof could mark a turning point in how science is conducted, signaling that AI may now operate as a co-investigator in the most abstract and complex domains. While the immediate practical impact on cryptography or digital security is likely minimal—since GRH does not provide a known method for factoring large integers—the broader societal effect may be more profound: a shift in the perceived boundaries of human intellectual capacity. The story echoes earlier moments of technological disruption, such as the 1997 defeat of chess grandmaster Garry Kasparov by IBM’s Deep Blue, but with far greater stakes, as it involves the creation of new knowledge rather than just pattern recognition. Institutions like universities, research labs, and funding bodies may soon face pressure to integrate AI into core research processes, potentially altering career trajectories and academic hierarchies. Meanwhile, concerns about AI autonomy and the potential for unverifiable or incomprehensible proofs—what some call “black box mathematics”—could intensify. The next phase will likely involve deeper scrutiny by the mathematical community, including formal verification and attempts to understand the AI’s reasoning process. Whether this leads to a new era of AI-driven science or raises ethical and epistemological questions about the nature of discovery remains to be seen.