Abstract: China's open-source large AI models, represented by DeepSeek, are contributing to bridging the global digital divide. Through a series of technological innovations, these models reduce the access divide by lowering costs; through more permissive open-source strategies, they lower usage barriers and mitigate the usage divide; and by narrowing disparities in digital literacy, they help reduce the capability divide. At the same time, they also give rise to new risks, including data sovereignty conflicts in the context of cross-border data flows, compatibility tensions between intellectual property rights and open-source licenses, and challenges arising from the fragmentation of global AI governance rules and the politicization of governance practices. Addressing these challenges requires enhanced international cooperation among stakeholders to promote a more equitable, inclusive, and effective global digital governance order. This includes establishing data classification and tiered governance rules, as well as regulatory sandbox mechanisms, to facilitate cross-border data flows and mitigate data sovereignty conflicts; accelerating intellectual property legislation for emerging technologies to improve international governance coordination; and promoting limited alignment between domestic and international legal frameworks to explore multilateral governance rules for artificial intelligence.