Anthropic Researcher Claims Claude AI Formulated Proof for Riemann Hypothesis
A mathematician at AI lab Anthropic claims that Claude helped produce a formal proof for the Riemann hypothesis, a 167-year-old problem long considered the holy grail of mathematics.

In this piece
An Anthropic researcher has claimed that the company's Claude AI model helped construct a formal proof for the 167-year-old Riemann hypothesis. Encoded in the…
An artificial intelligence researcher at Anthropic has made the extraordinary claim that the company's Claude model helped construct a formal proof for the Riemann hypothesis, widely considered the holy grail of unsolved mathematical problems. Mathematician Levent Alpöge announced that he spent roughly six weeks collaborating directly with the AI system to assemble the complex proof. Rather than presenting the result as a conventional academic paper, Alpöge released the work as thousands of lines of formal computer code.
First proposed by German mathematician Bernhard Riemann in 1859, the 167-year-old conjecture concerns the fundamental distribution of prime numbers along the critical line of the Riemann zeta function. Resolving the hypothesis has frustrated generations of leading thinkers, leading the Clay Mathematics Institute to designate it one of seven Millennium Prize Problems in 2000 with a one-million-dollar reward.
According to Alpöge, drafting the argument in standard mathematical prose would have produced an unreadable draft spanning hundreds of thousands of pages. To circumvent that logistical hurdle, the research team authored the proof entirely within Lean, an open-source formal verification language and interactive theorem prover. The platform requires every deduction to be validated mechanically from foundational axioms, preventing the subjective ambiguities that often plague human-written drafts.
The choice of Lean reflects a growing movement within pure mathematics toward computer-assisted formalization. In traditional publishing, reviewing a massive manuscript of this magnitude could consume years of committee labor and remain vulnerable to hidden logical oversights. By translating mathematical thought into machine-executable definitions, the contributors aimed to provide a transparent, reproducible framework for peer review.
Initial automated assessments have provided an intriguing early signal of legitimacy. External mathematicians, including Massachusetts Institute of Technology professor Andrew Sutherland, reported that the project's Lean files successfully compiled and passed through the Lean verification kernel using standard axioms. While compilation does not automatically guarantee full conceptual correctness, it confirms that the internal logical dependencies contain no basic syntax failures.
Despite that technical milestone, the wider mathematical establishment is proceeding with deep caution. Experts emphasize that verifying a formal proof requires confirming that the definitions entered into the code faithfully match the classical Riemann hypothesis without accidental simplifications or subtle mathematical loopholes. History is filled with premature announcements surrounding the problem, making independent line-by-line scrutiny indispensable.
If the proof is ultimately validated, the accomplishment would represent a watershed moment in the history of artificial intelligence and human knowledge. While frontier AI models have demonstrated proficiency in solving Olympiad-level exercises and summarizing complex literature, cracking an unsolved landmark conjecture would prove that machine learning systems can contribute meaningfully to novel scientific discovery.
The international verification process is expected to demand significant time and effort from specialists in both analytic number theory and computer science. Teams of researchers across global institutions are already downloading the repositories, attempting to deconstruct the logical architecture and independently verify each intermediate theorem.
Regardless of the ultimate verdict, the project highlights a fundamental shift in how mathematics may be practiced in the coming decades. The deep integration of human intuition and algorithmic theorem proving suggests that complex intellectual frontiers may increasingly be explored through collaborative human-machine partnerships.
Primary source: The Indian Express. Newzly rewrites and structures source reporting for clarity; verify critical details at the original.