China’s frontier AI safety field lacks funding, access, and risk awareness
Chinese AI safety researchers face a basic ecosystem problem: limited frontier-risk awareness, scarce domestic funding, and costly access to the best models make safety work hard to start and sustain. A field note based on interviews says some researchers sell safety products or courses to fund largely nonprofit research, while foreign outreach is constrained by fear of being seen as involving a foreign entity. It suggests supervised model subscriptions and compute for civilian safety researchers as an immediate remedy.