Jewitt、Rajbahadur、Li、Adams、Hassan（Queen’s University，KDD 2026）审计了 124,278 条 dataset→model→application 供应链，发现只有 2.3% 的宽松许可数据集和 3.2% 的宽松许可模型同时具备许可证全文与版权声明；而当上游真的齐备时，归属声明传到下游的比例只有 27.59%（dataset→model）与 5.75%（model→application）。结论一句话：宽松许可的标签不是可执行的授权，许可证文件与声明文本而非元数据，才是法律真相源。
Moonshot AI targets August 27 closing for its pre-IPO round ahead of a Hong Kong IPO filing, crystallizing the institutional transition from open-weight AI lab to listed entity; DeepSeek officially launches V4 Pro on August 13 at $0.435/M input tokens — the first frontier-class Chinese AI model priced for sustained revenue generation rather than cost-leadership signaling; The Guardian’s August 7 institutional analysis challenges China’s ‘open ecosystem’ narrative, arguing that China’s open-weight AI ecosystem is no more open than its US counterpart. This cycle’s briefing examines the institutional transition from subsidy-financed distribution to capital-market-financed licensing.
DeepSeek releases V4-Flash under MIT license — the first Chinese frontier open-weight model under a truly permissive OSI-style license; Alibaba moves Qwen toward revenue-sharing for large commercial users, creating a new ‘open weights, paid use’ institutional form; Reuters and Yahoo report Beijing has met with Alibaba and ByteDance to discuss curbing overseas access to its most advanced AI models, moving the July export-control debate from rumor toward implementation.
China launches WAICO, a 29-nation AI governance bloc backing open-source AI; Mulan License Family donated to OpenAtom Foundation; China’s anthropomorphic AI rules take effect; Beijing weighs export controls on its own AI models.
NDRC eight-point AI Action Plan, Trump admin considers ban on Chinese open-source models, Mulan copyleft license submitted to OSI, and Qiushi essay on AI ideology