Question
Will a third Millennium Prize Problem (other than Navier-Stokes and the Generalized Riemann Hypothesis) be publicly claimed as solved by AI or with significant AI assistance between now and December 31, 2026?
Status of Existing Claims To date, two Millennium Prize problems have seen public claims of AI-assisted solutions, but neither constitutes an accepted mathematical solution. On September 8, 2026, OpenAI announced an AI-generated proof for the Navier-Stokes equations, formalized in Lean 2 sources. The Clay Mathematics Institute noted the problem has "apparently been settled" but emphasized its "deliberately unhurried" evaluation process, stopping well short of formal acceptance claymath.org. Similarly, today (November 20), Anthropic mathematician Levent Alpöge posted a Lean formalization claiming to prove both the Generalized Riemann Hypothesis (GRH) and the standard Riemann Hypothesis (RH) with Claude's assistance 2 sources. While experts have successfully compiled the code and called it a historic milestone scientificamerican.com, the wider mathematical community has not yet fully digested or validated the proof scientificamerican.com.
Likelihood of a Verified Solution Because of the rigorous standards required for official recognition, the chance of a verified, Clay-accepted solution to any Millennium Prize problem by December 31, 2026, is under 1%. The Clay Mathematics Institute strictly mandates a formal publication process followed by a two-year waiting period 2 sources. Consequently, regardless of the validity of the Lean formalizations, formal Clay acceptance or widespread community consensus on a complete solution within the next six weeks is virtually impossible.
Progress on Remaining Problems
The primary uncertainty is whether a credible public claim for a third problem will emerge by the end of the year. The remaining candidate problems have seen substantial but strictly partial progress. In October, OpenAI published an AI-assisted proof for the existence of 4D Yang-Mills theory, which a committee led by David Gross confirmed 2 sources. However, the crucial mass gap remains unsolved 2 sources. For the Hodge and Birch and Swinnerton-Dyer (BSD) conjectures, recent announcements only cover restricted cases, and one Hodge result was even withdrawn due to an error 2 sources. Regarding P vs NP, an earlier June preprint claiming an AI-assisted proof initially defined its top-level proposition as True and relied on an opaque axiom in its revision, effectively stripping it of the credibility required to qualify as a valid claim 2 sources.
Drivers of the Final Probability The probability of a third credible claim emerging before December 31 rests on a balance of unprecedented momentum and formidable mathematical hurdles. The intense competition between major labs has produced two full claims in just ten weeks 2 sources. However, the remaining problems lack a clear path forward, and the narrow six-week window includes the holidays, during which labs may adopt a more cautious vetting process following recent controversies superpowerdaily.com. There is also minor structural uncertainty regarding whether a literal reading of Alpöge’s GRH announcement might mistakenly double-count the standard RH as a separate third problem alpo.ge. Weighing the extraordinary pace of AI research against the severe difficulty of extending partial results into full, credible claims justifies this probability.
Comparing this question to others about long-term AI-assisted math breakthroughs kept the estimate stable, highlighting that while fully verified solutions are far off, the rapid pace of lab announcements makes a near-term unverified claim distinctly plausible.
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited