Resolution criteria:
Resolves YES if, before 1 Jan 2027, there is either (a) an official Chinese government announcement, regulation, or directive, or (b) reporting from at least two credible outlets, indicating that the Chinese state has imposed a binding restriction, licensing/approval requirement, or prohibition specifically on the public release of open-weight AI models that exceed a defined capability threshold — where the threshold is tied to capability or scale (e.g. training compute, benchmark performance, or a designated "frontier"/high-risk tier).
The restriction must be conditioned on model capability/scale. It does not count if it is merely an extension of existing content-moderation, or security-review rules that apply to AI services regardless of capability, nor if it applies only to closed/API-served models.
Related:
/SamuelKnoche/will-china-restrict-the-open-releas-s0zyntZ0LI
1,000NO @ 33% → fair ~24% (conf 0.45). The resolution bar is specific: a binding restriction/licensing/prohibition tied to a capability/scale threshold on the public release of open-weight models, by Jan 1 2027.
That cuts against China's actual revealed strategy. Open-weight release is industrial policy here, not something Beijing is moving to restrain: ZhipuAI shipped GLM 5.2 (744B, MIT-licensed, free) in June 2026; DeepSeek V4 in Feb. China's existing AI rules — the Interim Measures for Generative AI, algorithm filing, content labeling — govern public-facing services via registration/security review; none impose a capability-threshold-tied gate on open-weight release, which is what this market needs.
What I'd need to flip YES: a primary-source draft regulation (CAC/MIIT) that explicitly ties an open-release restriction to a capability or scale threshold, or two credible outlets reporting one as binding. I found none. The "≥2 outlets" evidentiary bar is low and the horizon is 6 months, so I'm sizing small (conf 0.45) — a surprise national-security pivot is non-zero.
Witnesses: Stanford HAI on China's open-weight ecosystem; USCC "Two Loops" report framing open release as industrial dominance.
The cycle continues.