Blab AI
OpenAI Claims AI-Generated Proof of Navier-Stokes Blowup Sparks Credit Dispute
On September 8 2026, OpenAI announced that one of its internal AI systems had produced a proof claiming that the Navier–Stokes equations, which govern viscous fluid motion, can develop a singularity in finite time. The 166‑page manuscript was bolstered by a formal verification run in the Lean proof assistant.
The Navier–Stokes existence and smoothness problem is one of the seven Millennium Prize Problems identified by the Clay Mathematics Institute in 2000; a correct solution would earn a $1 million prize. OpenAI, however, stated that it would not pursue the award.
The proof tackles the forced version of the equations, where an external force drives the fluid. The AI identified a vortex that thins and accelerates until its speed diverges, thereby exhibiting a blowup. The effort involved a network of roughly 10,000 concurrent AI agents running for about 88 hours, with OpenAI estimating the cost at several million dollars.
The announcement follows a related advance by mathematicians Tristan Buckmaster of New York University and Levent Alpöge of Anthropic. Earlier in September, Buckmaster and Alpöge published a forced‑Euler blowup result that also leveraged AI models—including OpenAI’s GPT‑4. Since the Euler equations are a simplified, inviscid counterpart of Navier–Stokes, the forced‑Euler blowup served as a stepping stone toward the more complex Navier–Stokes problem.
The close timing of the two announcements has sparked speculation about their interconnection. Buckmaster stated that he and Alpöge had spoken with OpenAI researchers about how to present the results, alleging that the discussions included a request to exclude Alpöge from the publication because he works for Anthropic, a competitor. He added that the community deserves a “serious and unhurried discussion” about the implications.
OpenAI denied that its AI agents accessed the mathematicians’ research directly. The company clarified, “while unlikely, we cannot rule out that de‑identified data derived from their usage of our products helped improve our models.”
While the Lean verification has confirmed the proof’s internal consistency, the broader mathematical community has yet to conduct a comprehensive human review. Gregory Eyink, a mathematical physicist at Johns Hopkins University, remarked, “I don’t think anyone has completely verified the proof yet, certainly not on the human side.”
The result is primarily of theoretical interest. While the Navier–Stokes equations treat fluids as continuous media, real fluids consist of discrete molecules, causing the equations to break down at very small scales. Eyink noted that the practical impact is limited, but the prestige of solving a Millennium Prize Problem remains significant.
The announcement fits into a broader trend of AI‑assisted breakthroughs in mathematics over the past year. OpenAI’s earlier work on an Erdős problem in 2023, coupled with the current Navier–Stokes claim, showcases the increasing influence of large language models and multi‑agent systems in addressing complex mathematical challenges.
The credit dispute underscores a new challenge: how to attribute discoveries when AI systems play a role in the reasoning process. Eyink described the issue as “the really serious ongoing problem and issue that has to be resolved.”
At present, the Navier–Stokes proof remains unverified by external mathematicians, and the Clay Mathematics Institute has not issued an official statement. The academic community is expected to continue scrutinizing the 166‑page manuscript and its Lean formalization.
The controversy also highlights the need for clear guidelines governing the use of AI in research and the sharing of data that may influence model development. Whether OpenAI’s claim will ultimately be accepted as a solution to the Millennium Prize Problem remains to be seen.
The Navier–Stokes existence and smoothness problem is one of the seven Millennium Prize Problems identified by the Clay Mathematics Institute in 2000; a correct solution would earn a $1 million prize. OpenAI, however, stated that it would not pursue the award.
The proof tackles the forced version of the equations, where an external force drives the fluid. The AI identified a vortex that thins and accelerates until its speed diverges, thereby exhibiting a blowup. The effort involved a network of roughly 10,000 concurrent AI agents running for about 88 hours, with OpenAI estimating the cost at several million dollars.
The announcement follows a related advance by mathematicians Tristan Buckmaster of New York University and Levent Alpöge of Anthropic. Earlier in September, Buckmaster and Alpöge published a forced‑Euler blowup result that also leveraged AI models—including OpenAI’s GPT‑4. Since the Euler equations are a simplified, inviscid counterpart of Navier–Stokes, the forced‑Euler blowup served as a stepping stone toward the more complex Navier–Stokes problem.
The close timing of the two announcements has sparked speculation about their interconnection. Buckmaster stated that he and Alpöge had spoken with OpenAI researchers about how to present the results, alleging that the discussions included a request to exclude Alpöge from the publication because he works for Anthropic, a competitor. He added that the community deserves a “serious and unhurried discussion” about the implications.
OpenAI denied that its AI agents accessed the mathematicians’ research directly. The company clarified, “while unlikely, we cannot rule out that de‑identified data derived from their usage of our products helped improve our models.”
While the Lean verification has confirmed the proof’s internal consistency, the broader mathematical community has yet to conduct a comprehensive human review. Gregory Eyink, a mathematical physicist at Johns Hopkins University, remarked, “I don’t think anyone has completely verified the proof yet, certainly not on the human side.”
The result is primarily of theoretical interest. While the Navier–Stokes equations treat fluids as continuous media, real fluids consist of discrete molecules, causing the equations to break down at very small scales. Eyink noted that the practical impact is limited, but the prestige of solving a Millennium Prize Problem remains significant.
The announcement fits into a broader trend of AI‑assisted breakthroughs in mathematics over the past year. OpenAI’s earlier work on an Erdős problem in 2023, coupled with the current Navier–Stokes claim, showcases the increasing influence of large language models and multi‑agent systems in addressing complex mathematical challenges.
The credit dispute underscores a new challenge: how to attribute discoveries when AI systems play a role in the reasoning process. Eyink described the issue as “the really serious ongoing problem and issue that has to be resolved.”
At present, the Navier–Stokes proof remains unverified by external mathematicians, and the Clay Mathematics Institute has not issued an official statement. The academic community is expected to continue scrutinizing the 166‑page manuscript and its Lean formalization.
The controversy also highlights the need for clear guidelines governing the use of AI in research and the sharing of data that may influence model development. Whether OpenAI’s claim will ultimately be accepted as a solution to the Millennium Prize Problem remains to be seen.
