Anthropic's Unreleased AI Model Advances Riemann Hypothesis Verification to 67%

Anthropic said August 10, 2026 that an unreleased AI model lifted the verified share of Riemann zeta zeros from 41.6% to 67.2% — the biggest jump in decades on the 150-year-old problem. The result was formalized in Lean, a proof-checking language, and confirmed by two staff mathematicians. It extends a lower bound, not a full proof. The Clay Institute's $1 million prize remains unclaimed.