OpenAI says its AI agents have solved one of the most important open problems in mathematics, a 90-year-old puzzle surrounding the Navier-Stokes equations that has frustrated generations of the world’s best minds. Under ordinary circumstances, such a breakthrough would be a historic milestone. But the announcement has been overshadowed by accusations that OpenAI failed to credit researchers whose AI-assisted work influenced its solution.
Whether those accusations are true or not, the episode marks a turning point in the history of mathematics. AI models increasingly appear essential to making progress on the field’s hardest problems, yet the scale of computing resources required to achieve such results may be available only to a couple of frontier AI companies. If that is the future mathematics is headed toward, the role of the human mathematician will need to be fundamentally redefined.
The 90-Year-Old Math Problem OpenAI Says It Solved
The Navier-Stokes equations are the mathematical backbone of fluid dynamics. They describe how liquids and gases move, the flow of air over an airplane wing, the swirl of water down a drain, the currents that drive the planet’s oceans and atmosphere. Since the 1930s, mathematicians have struggled with a deceptively simple question: do these equations always have smooth, physically realistic solutions in three dimensions, or can they break down in ways that defy physical sense?
What is the Navier-Stokes existence and smoothness problem?
The Navier-Stokes existence and smoothness problem asks whether the equations governing fluid motion always produce solutions that remain smooth and physically meaningful in three dimensions. It was first posed in the 1930s, became one of the seven Millennium Prize Problems in 2000, and carries a $1 million prize for a correct solution.
OpenAI says it attacked the problem with 10,000 AI agents working in parallel. According to the company, the agents produced a result in 88 hours. OpenAI describes the achievement as the first major mathematical problem solved by AI.
The effort was not a money-maker. OpenAI reportedly spent millions of dollars on computing resources to secure the $1 million reward, a reminder that frontier mathematical discovery is becoming an industrial-scale operation.
The Credit Controversy, Explained
Instead of celebrating, the research community is arguing over attribution. OpenAI stands accused of failing to credit researchers whose AI-assisted work influenced the solution. The specific details of the dispute are still emerging, but the fundamental question is one that mathematics has never had to confront at this scale: when an AI system solves a problem, whose work is it?
The dispute cuts to the heart of how scientific credit operates. In traditional mathematics, every proof is built on the published work of predecessors, and citation is the currency of the field. But AI systems train on vast bodies of existing work, including the unpublished and partially published work of other researchers, without any mechanism for attribution. When a frontier lab’s model produces a breakthrough, the contributions of those whose work shaped the model’s capabilities can simply disappear.
The Navier-Stokes episode is the first major test of this problem, and the mathematical community does not yet have an agreed answer.
A Deep Blue–Kasparov Moment for Mathematics
NYU mathematician Tristan Buckmaster has compared the event to one of the most famous milestones in computing history: the day in 1997 when IBM’s Deep Blue supercomputer defeated chess world champion Garry Kasparov. “This is a Deep Blue–Kasparov moment,” Buckmaster wrote in a statement. “The community needs to have serious and unhurried discussion about where to go from here.”
Deep Blue’s victory was a shock, but its implications were narrow. The machine could play chess and little else. The AI systems now proving themselves in mathematics are different. They are general-purpose, rapidly improving, and concentrated in the hands of a few corporations. That changes the incentives in a field that has long been one of the most democratic intellectual pursuits in human history.
For centuries, a lone researcher with a notebook and an original idea could reshape mathematics. The next generation of breakthroughs may require access to millions of dollars in cloud computing and proprietary AI tools. If that becomes the baseline for frontier research, mathematics will no longer be a field open to anyone with talent and tenacity. It will be a field where a handful of companies set the agenda, own the results, and decide who gets to participate.
AI Accountability Is Under Strain Across the Industry
The OpenAI controversy is not occurring in a vacuum. The same week brought news of intensifying accountability problems across the AI sector.
The United States accused six Chinese AI firms, including DeepSeek, Moonshot AI, Alibaba, and Z.AI, of “industrial-scale” theft of American AI technology. The firms allegedly used a technique called model distillation to train their systems, drawing on outputs from major American models such as Claude, ChatGPT, Gemini, and Grok. Distillation, which uses a larger model’s outputs to train a smaller one, is highly efficient when done legitimately, but US officials say the practice in this case amounted to systematic theft of trade secrets.
The Pentagon, meanwhile, asked OpenAI to build an AI model that rarely says no. The military wanted a system with “minimal refusal rates.” The request raises questions that will only become more urgent as AI is integrated into military decision-making.
Meta also launched a new AI agent called Muse, which autonomously accesses apps and websites to send emails and make payments on a user’s behalf. Internal tests, however, found that the system could expose sensitive personal data. The launch is a reminder that even mainstream consumer AI products carry risks that are not fully understood until they are deployed.
The Week’s Other Defining Tech Stories
US battery installations hit another record. New capacity in the second quarter of 2026 reached 20.2 gigawatt-hours, enough to supply the daily electricity needs of 600,000 homes. Cheaper batteries and the urgency of storing renewable power are driving the surge, though grid-scale and residential markets are evolving along different paths.
Apple is expected to unveil a $2,000 folding iPhone this week. The device would be the biggest design change to the iPhone since its 2007 launch and the first major test for new CEO John Ternus. Xiaomi and Huawei both released new foldables in the days before Apple’s event, signaling a crowded market.
Google says EU compliance is degrading its search results. The company says it must give more prominence to comparison sites in Europe to avoid further penalties, following a €460 million antitrust fine. Google’s own assessment is that the change lowers the quality of results for European users.
Border Patrol is using financial data to target Americans for stops. A predictive-policing program at the Department of Homeland Security analyzes US citizens’ financial habits and feeds intelligence to local police, who use it to decide who gets stopped.
New paints can cool buildings without electricity. The coatings reflect sunlight and radiate heat back into space, offering a low-cost way to reduce building temperatures in a warming climate.
The first trailer for the Sam Altman biopic has arrived. Luca Guadagnino’s Artificial, starring Andrew Garfield, opens in US theaters on December 25. Amazon dropped the film after investing in OpenAI, a decision that speaks to the uneasy relationship between Hollywood, tech money, and AI.
Danijar Hafner is developing AI agents that plan ahead. The 31-year-old entrepreneur, one of the artificial intelligence honorees on the 35 Innovators Under 35 list for 2026, trains agents to navigate environments they have never encountered during training. His San Francisco office is filled with humanoid robots, and he is beginning to move his agents from virtual worlds into physical reality.
Ukraine’s drone data is fueling a new Wild West marketplace. Millions of data points gathered from tens of thousands of drone flights are being made available to military contractors and commercial companies. The data is immensely valuable for training AI systems, but it turns the battlefield into a training ground, prompting calls for a regulatory framework that treats conflict data as more than ordinary commercial material.
The Unseen Human Toll of the AI Boom
While mathematicians debate the future of their discipline, another, less visible consequence of AI’s expansion is unfolding in the world of adult content. AI systems are training on the work of adult performers, cloning their likenesses, and generating explicit content they never consented to create. The people affected have little legal protection and even less control.
Take the case of Jennifer, a performer who, in 2023, ran a new professional headshot through a facial recognition program. She wanted to see whether it would find the porn videos she had made more than a decade earlier. It did, but it also found something else: one of her old videos, now featuring someone else’s face on her body.
Discussions of sexualized deepfakes generally focus on the people whose faces are inserted into explicit content without consent. But there is another group that is almost entirely overlooked: the people whose bodies those faces are attached to. They suffer the same violation of agency and the same damage to their livelihoods, but without the same legal recognition or public sympathy.
This is the darker backdrop to the celebration of AI’s mathematical achievements. A technology capable of solving the Navier-Stokes equations is also capable of destroying a person’s sense of ownership over their own body and work. That contradiction will have to be addressed if the AI era is to deliver on its promise.
A Milestone That Calls for More Than Celebration
The solution to the Navier-Stokes problem, if it withstands verification, will be remembered for decades. But the circumstances surrounding its announcement, the credit dispute, the concentration of resources, the broader accountability failures across the industry, will be remembered too.
Tristan Buckmaster’s call for a “serious and unhurried discussion” was not only about mathematics. It must also cover the researchers whose work is absorbed into AI systems without acknowledgment, the performers whose bodies are cloned without consent, the companies accused of training on others’ intellectual property, and the publics who are being asked to trust systems that remain opaque to them.
The Deep Blue–Kasparov moment transformed public understanding of what machines could do in a closed, rule-bound game. The Navier-Stokes moment raises a more consequential question: in a world where machines can do the work of the human mind, who will own that capability, and who will benefit from it? The community’s serious and unhurried discussion is overdue. The question now is whether it happens before the next milestone, and before the next person discovers that an AI has used their work, their body, or their ideas without asking.
