A mathematical proof produced with an internal OpenAI language model has intensified debate over what counts as solving the Navier–Stokes millennium problem. The work appears to satisfy an option in the formal statement published by the Clay Mathematics Institute, but researchers quoted in a September 22 report say it does not resolve the unforced version that has long driven the field.
Navier–Stokes equations describe fluid motion. The million-dollar problem asks, in part, whether smooth solutions can develop a singularity—a mathematical blow-up in which quantities become unbounded. Much of the research effort focuses on whether such a failure can emerge from a fluid's intrinsic dynamics, without an engineered external force.
OpenAI's construction takes a different route. It uses a specifically designed force to produce a blow-up. That approach is formally relevant because the institute's original problem statement, written by mathematician Charles Fefferman in 2000, includes an option allowing an external force. On that reading, the proof addresses an explicitly permitted formulation rather than changing the rules after the fact.
The distinction matters scientifically. Researchers described the applied force as highly contrived and unlike a force expected in a realistic fluid. Luis Silvestre of the University of Chicago told Scientific American that the Clay formulation may be settled even though what he called the most important version remains unsolved. The result therefore separates eligibility under the written challenge from the broader question mathematicians intended the challenge to illuminate.
The controversy sharpened after three mathematicians posted another proof showing that OpenAI's method cannot simply be extended by removing the external force. According to the report, the blow-up disappears without it. If that analysis holds, the construction is not an intermediate step that can be gradually converted into a solution of the unforced case; a substantially different idea would be required.
The episode also has a complicated research history. Diego Córdoba and Luis Martínez-Zoroa had developed a program for constructing forces that trigger singular behavior. Two other mathematicians then used that line of work, together with AI, to obtain a blow-up for a frictionless fluid on September 7. OpenAI reported its Navier–Stokes result less than a day later, and the timing generated a separate dispute over the company's role and credit.
No consensus described in the report erases the formal significance of the new proof. It demonstrates a way for a forced Navier–Stokes system to fail and exposes how an optional clause in a famous problem can shape the meaning of a claimed solution. It also gives mathematicians a clearer boundary: this particular mechanism depends on the engineered force.
The remaining challenge is the one most fluid dynamicists recognize instinctively—establishing regularity or finding a blow-up when the fluid is not driven by a specially chosen external input. OpenAI's result may satisfy the literal statement's force-enabled branch, but it has not answered that deeper unforced question.
