OpenAI Claims Navier-Stokes Solution Amidst IP Controversy

What Happened

On Tuesday, September 9, 2026, OpenAI announced it had developed a solution to the 90-year-old Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The company stated that its internal, unreleased AI model, described as "significantly more capable than GPT-6 Astra," utilized approximately 10,000 concurrent AI agents to arrive at a proof. This resolution was reportedly achieved on Saturday, September 5, after 88 hours of continuous operation, following the launch of the first agents on September 1. The AI system, which had access to tools like a cached internet version and code execution, exchanged about 2.7 million messages and processed roughly 130 billion output tokens during its effort. The proof was subsequently verified in the Lean proof language over an additional 17 hours. However, this announcement was immediately complicated by a public statement from Tristan Buckmaster, a mathematics professor at New York University, who alleged that OpenAI's research path closely mirrored his own unpublished work, which he had been developing with collaborator Levent Alpöge, an Anthropic mathematician, using OpenAI's own Codex tool.

What the Evidence Establishes

The evidence establishes that OpenAI initiated its effort to solve the Navier-Stokes problem on September 1, 2026, after hearing a rumor about progress on the puzzle. Their AI system, comprising 10,000 agents, produced a resolution on September 5, taking 88 hours. The company claims the agents used tools such as internet access and code execution, communicating within groups. The total compute cost for the wider effort is estimated between $10 million and $22.5 million. The Navier-Stokes equations describe fluid movement and are critical for fields like aircraft design and weather forecasting. The Clay Mathematics Institute, which established the Millennium Prize Problems in 2000, offers a $1 million prize for each solution. Separately, Professor Tristan Buckmaster and Levent Alpöge had been working on related problems, including the Euler equations (a frictionless variant of Navier-Stokes), proving their 'blow up' on August 15. They utilized OpenAI's Codex and Anthropic's Claude in their research, employing a technique called 'forcing' that Buckmaster states was not widely pursued.

Where the Accounts Conflict

A primary conflict exists regarding the origin of OpenAI's research direction and potential access to external work. Professor Buckmaster stated that he received "tips" indicating information about his and Alpöge's progress had been passed to OpenAI, and he found OpenAI's route to the solution similar to their own. He explicitly questioned whether OpenAI models had been trained on, or had access to, their Codex sessions. Buckmaster also alleged he was offered sole authorship on a paper that excluded Alpöge's name and was warned, "Why would you ruin your career?" when he threatened to go public. OpenAI, conversely, stated that its researchers and agents "did not see any of their work through any means until they released it publicly." The company explicitly denied accessing "specific user data" to solve the problem. However, OpenAI conceded, "While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models." Sam Altman, OpenAI's CEO, commented that the two approaches appear different now that both are public. Sébastien Bubeck, an OpenAI researcher, denies asking for Alpöge's name to be removed from any potential publication.

Context and Stakes

The Navier-Stokes problem is one of the most significant unsolved mathematical challenges, carrying a $1 million prize and profound implications for understanding fluid dynamics, from ocean currents to blood flow. A verified solution would represent a monumental scientific achievement. However, the controversy surrounding OpenAI's claim introduces substantial ethical and intellectual property stakes for the rapidly evolving field of artificial intelligence. The allegations by Professor Buckmaster highlight concerns about the proprietary nature of AI tools like Codex and Claude, and whether data fed into these systems by users could inadvertently or directly contribute to the AI company's own research, potentially undermining independent academic work. This situation raises questions about data privacy, research integrity, and the fair attribution of intellectual contributions in an era where AI models are increasingly powerful and opaque. As mathematician Terence Tao noted, the wider cost of such incidents is significant, likening it to "having machines that can lift weights for you at the gym," implying a shift in the nature of intellectual labor and its recognition.

What to Watch Next

The immediate next step involves the Clay Mathematics Institute of Cambridge, Massachusetts, which has not yet commented on OpenAI's proposed solution. The Institute will need to formally receive and verify the proof, a process that typically involves rigorous peer review by the mathematical community. This verification process is crucial, especially given that some mathematicians point out OpenAI's version leans on a term often excluded from the problem statement. Observers will also monitor for any further public statements or formal actions from Professor Buckmaster and Levent Alpöge, such as a detailed public release of their own work or a formal complaint regarding intellectual property. OpenAI's response to these ongoing allegations, particularly regarding its data usage policies and how it addresses concerns about researchers using its proprietary tools, will be closely scrutinized. Any clarification or policy changes from OpenAI regarding the use of user data for model training and internal research will be significant for the broader AI research community. The compute bill for the wider effort, estimated at $10 million to $22.5 million, also highlights the financial scale of such endeavors.

Bottom Line

OpenAI's announcement of a solution to the Navier-Stokes problem, a 90-year-old mathematical challenge, represents a potentially groundbreaking achievement for artificial intelligence. The company's use of 10,000 AI agents and an advanced internal model to achieve this in 88 hours underscores the accelerating capabilities of AI. However, this scientific claim is significantly overshadowed by serious allegations of intellectual property infringement and research misconduct from Professor Tristan Buckmaster and Levent Alpöge. The core dispute centers on whether OpenAI's AI models or researchers gained an unfair advantage by accessing or being influenced by the unpublished work of Buckmaster and Alpöge, who had used OpenAI's own Codex tool. This incident highlights a critical tension between rapid AI development and the ethical frameworks necessary to ensure fair play, transparency, and proper attribution in scientific research. The resolution of this controversy, particularly the Clay Mathematics Institute's verification and any further actions regarding the IP claims, will have lasting implications for the future of AI research ethics and the relationship between AI companies and the academic community.