Last Week in AI #344 - Navier–Stokes, Pacing the Frontier, AI Misuse

OpenAI announced that its internal GPT‑6 Astra–based agent system produced a proof of the Navier–Stokes existence and smoothness problem, claiming a solution to one of the Clay Millennium Prize problems by demonstrating finite‑time vortex blowup. The claim triggered sharp criticism from mathematicians—including NYU’s Tristan Buckmaster—who question its validity and note the lack of public peer review, though OpenAI has provided a Lean formalization for verification. In response to concerns about rapid AI progress, Anthropic’s CEO called for pacing development, and regulatory discussions are intensifying amid warnings that unchecked AI capabilities could pose ecological risks.

Source: Last Week in AI | 2026-09-17