Anthropic Researcher Claims Claude Proved Riemann Hypothesis—But No Official Confirmation

Anthropic researcher Levent Alpöge has claimed that the company’s Claude model helped him prove the Riemann hypothesis, a longstanding problem in mathematics considered the most important unsolved question in the field. Alpöge announced the discovery on his personal website today, posting a four-page statement describing a "proof" of both the standard and generalized versions of the conjecture—one that, according to his account, spans over 250,000 pages. Anthropic, which went public on the Nasdaq just one day earlier in a $93 billion valuation, has not issued an official response.
The claim, if verified, would mark a historic milestone for both AI and mathematics. The Riemann hypothesis, proposed by Bernhard Riemann in the mid-1800s, concerns the distribution of prime numbers and has implications across number theory, cryptography, and physics. Alpöge’s assertion—that Claude not only solved the problem but also a broader "generalized" variant—would position Anthropic’s models as capable of tackling foundational scientific challenges, far beyond current benchmarks in AI-assisted research.
Alpöge’s evidence includes a Lean-formalized version of the proof, a programming language used to verify mathematical results with near-certainty. Lean proofs are treated as software programs: if the code runs without errors, the mathematical argument is considered correct. His Lean file reportedly contains millions of lines of code, a scale that suggests the proof’s complexity far exceeds anything previously attempted. Yet the full 250,000-page "natural language" proof remains unpublished, leaving the mathematical community skeptical—or, at minimum, deeply cautious.
The timing of the announcement is striking. Anthropic’s Nasdaq debut, which followed speculation about its ability to compete with major rivals, came just hours before Alpöge’s claim surfaced. While the IPO itself was framed as a validation of the company’s safety-focused AI research, this new development could redefine its market positioning. If true, it would demonstrate that Claude’s capabilities extend beyond chatbot-like interactions into the realm of *generative discovery*—a leap that could accelerate Anthropic’s adoption in academia, defense, and high-stakes scientific research.
But skepticism is warranted. No peer-reviewed journal has examined the work, and Alpöge’s claim lacks the endorsement of Anthropic’s leadership or independent verification. Alpöge’s work, if conducted under those terms, would have required approval—a detail that raises questions about why the company has remained silent. Alternatively, if he acted independently, it could signal gaps in Anthropic’s oversight of its own tools.
The broader implications for AI research are equally profound. If Claude can generate proofs of this scale and rigor, it would suggest that current benchmarks—such as the ability to pass math competitions or write functional code—understate the technology’s potential. This could pressure competitors to invest more aggressively in mathematical reasoning, potentially diverting resources from other areas like multimodal or creative AI. For Anthropic, a company that has emphasized *interpretability* and *safety* over raw performance, a verified proof could serve as a powerful differentiator in a crowded market.
Yet the absence of official confirmation from Anthropic is telling. The company’s IPO filings and public statements have consistently framed its work as incremental, focused on iterative improvements to safety and alignment. A claim of this magnitude—especially one that could attract scrutiny from mathematicians, regulators, and competitors—would typically warrant immediate clarification. The silence may reflect internal debates over how to handle an unvetted, potentially world-changing assertion, or it could indicate that Anthropic’s leadership is still assessing the validity of Alpöge’s work.
What’s next? The mathematical community will likely demand access to the full proof, particularly the Lean formalization, before taking the claim seriously. Independent researchers may attempt to replicate or audit the results. Anthropic’s response—or lack thereof—will also be closely watched. If the company chooses not to engage, it risks eroding trust among its core audience: investors, researchers, and policymakers who prioritize rigor over hype.
For now, the story remains a tantalizing "what if." If Alpöge’s claim holds up, it could mark the first time an AI system has contributed to a proof of such fundamental importance. If it doesn’t, it serves as a cautionary tale about the risks of unverified claims in an era where AI’s capabilities are both celebrated and scrutinized. Either way, the episode highlights the high stakes of Anthropic’s moment—and the challenges of balancing breakthrough potential with responsible disclosure in the race to define the future of intelligent machines.
Sources: siliconangle.com
“This claim, if true, would redefine AI’s role in scientific discovery—and Anthropic’s standing as a leader in mathematical reasoning—just days after its $93B IPO.”
Read the original reporting
The outlets below did the original reporting.
- Anthropic researcher claims to have proven Riemann hypothesis using Claude — siliconangle.com
Related briefs
- Dutch Electric Aviation Startup Elysian Aircraft Raises €12M to Expand Civil and Defense A
- Science’s $400M Brain-Computer Interface Round Signals Neurotech’s Breakout Moment
- Spotify’s Daniel Ek backs AI researcher’s fusion bet—this time, simulations come first
- MegaRobo’s $200M Haul Leads Asia’s $335M Funding Week as Semiconductors and AI Infrastruct
- Archer Raises $10M to Fix AI’s Broken Hiring Arms Race—With a Bipartisan Platform
This brief was drafted automatically from the sources above and published under our editorial policy. Spotted an error? Tell us.