What happened
Update: Carnegie China released version 3.0 of its Global AI Talent Tracker on 23 September 2026, measuring authorship at NeurIPS, ICML and ICLR (undergraduate institution as a proxy for country of origin, current affiliation for location). The headline finding: China now hosts 40.6% of the world's top AI researchers, up from 27.1% in 2022, and '57% of top AI talent originate from China, up from 46% in 2022.' Chinese-origin AI talent migration to the US has fallen 89% since 2017, while 87% of Chinese-origin researchers already established in the US stayed. The report also flags an 'invisible wall' inside China: fewer than 15% of Chinese university AI researchers have industry ties, versus 37% at Carnegie Mellon. The 3.0 update automated parts of the data work with Claude Code and OpenAI Codex under human quality control (data/update note; not a peer-reviewed paper).
Why it matters
The tracker disrupts the assumption that the US still leads the global AI talent race on affinities that matter most — boards and talent leaders should treat China's research base as the binding talent constraint, not just its compute access.
Action needed
Re-baseline AI talent and recruiting strategy against the tracker's per-country retention data; monitor China's academia-industry tie gap as a leading indicator of commercialization acceleration.