OpenAI announced on September 8 that an advanced internal model has resolved the Navier–Stokes Millennium Prize Problem, one of mathematics' most celebrated unsolved challenges carrying a $1 million prize from the Clay Mathematics Institute. The achievement reportedly consumed 300 billion output tokens valued at $22.5 million, with approximately 10,000 concurrent AI agents working for roughly 88 hours. However, the announcement was immediately overshadowed by a serious credit dispute with NYU mathematician Tristan Buckmaster, who alleged that OpenAI obtained knowledge of his concurrent research on a related problem and inappropriately pressured him to remove co-author attribution to Anthropic's Levent Alpöge. While OpenAI denied the allegations and announced it would not claim the prize, the controversy drew sharp warnings from 25 Fields medalists about rushed announcements raising "severe attribution and plagiarism questions."
Meanwhile, Anthropic CEO Dario Amodei published a policy proposal titled "Pace the Frontier," calling for deliberate slowing of AI development to address mounting safety concerns. Amodei outlined three strategies: embedding independent third-party evaluators from organizations like METR within AI companies to monitor safety commitments and ensure incident reporting; coordinating among leading companies on common safety standards and limits on development rates; and pursuing global coordination, including with China, on prohibitions of narrow AI misuses such as biological weapons production. The proposal directly calls on governments to require frontier companies to match Anthropic's commitments and to mediate industry coordination talks—marking a significant shift toward regulatory oversight in the AI sector.
The two developments underscore deepening tensions between rapid capability advancement and calls for safety-first governance. Anthropic researchers have publicly raised alarms about current development practices, contributing to departures from the company and Google over safety concerns. Together, these stories reflect an inflection point in the AI industry, where technical breakthroughs are colliding with intensifying demands for slower, more accountable development practices.
Key Points
OpenAI claims resolution of the Navier–Stokes Millennium Prize Problem with an internal model, but faces serious allegations that it leveraged knowledge of competitors' concurrent research
25 Fields medalists issued an open letter warning against rushed announcements and attribution practices, with the Clay Institute declining to accept the proof pending scrutiny
Anthropic CEO Dario Amodei proposes "pacing the frontier" through embedded independent evaluators, industry coordination on safety standards, and government-mediated global agreements on AI development rates
Researchers at Anthropic and Google are departing over safety concerns, signaling internal disagreement on appropriate development speed and risk management
The episode highlights growing friction between AI companies' race for breakthroughs and emerging consensus among researchers and policymakers for more deliberate, safety-focused development