1、State of AI Safety in China(2025)AuthorsThis report was authored by Gabriel Wagner,Jason Zhou,Kwan Yee Ng,and Brian Tse.AcknowledgementsWe are grateful to Alan Chan,Xin CHEN,Jeff Ding,Oliver Guest,Emmie Hine,Lucy Luo,Zilan QIAN,SaadSiddiqui,Jan Pieter Snoeij,and Jie ZHANG for providing feedback and

2、suggestions on this report.How to cite this reportGabriel Wagner,Jason Zhou,Kwan Yee Ng,and Brian Tse,“State of AI Safety in China(2025),”ConcordiaAI,July 2025,https:/concordia- of ContentsExecutive Summary1Introduction and Scope41Domestic Governance71.1Overarching national guidance.71.2Legislation

3、and regulation.111.3Standards.141.4Government-backed AI safety and governance institutions.212International Governance222.1Overarching features of Chinas international AI governance.222.2Chinese actions at the United Nations.242.3Bilateral engagements.252.4Multilateral engagements.272.5Track 1.5 and

4、 2 dialogues.293Technical Safety Research363.1Methodology.363.2Overall trends.373.3Relevant research groups.383.4Notable research contributions.414Expert Views on AI Safety and Governance444.1Discourse trends at top AI conferences.454.2Discourse trends for key AI safety topics.495Industry Governance

5、535.1Collective industry action.545.2Individual company action.555.3Safety as a service.62iiTable of ContentsState of AI Safety in China(2025)Conclusion65Acronyms67Notes71iiiExecutive SummaryAs AI capabilities and governance challenges continue to growand Chinese models close their gap with theleadi

6、ng edgeunderstanding Chinas role in AI safety and governance is more critical than ever.Since 2023,Concordia AIs annual State of AI Safety in China reports have analyzed how China addresses general-purposeAI risks,particularly dangerous misuse,accidents,and loss of controlchallenges with an outsized