AI治理国际评估指数（AGILE Index 2025）

AGILEIndex2025 (ReleasedinJuly2025) CenterforLong-termArtificialIntelligence(CLAI)BeijingInstituteofAISafetyandGovernance(Beijing-AISI)BeijingKeyLaboratoryofSafeAIandSuperalignmentInternationalResearchCenterforAIEthicsandGovernance,InstituteofAutomation,ChineseAcademyofSciences CitationInformation: YiZeng,EnmengLu,XiaoyangGuo,CunqingHuangfu,JiaweiXie,YuChen,ZhengqiWang,DongqiLiang,GongceCao,JinWang,ZizheRuan,XinGuan&AmmarYounas.(2025).“AIGovernanceInternationaLEvaluationIndex(AGILEIndex)2025”.Beijing:CenterforLong-termArtificialIntelligence(CLAI);BeijingInstituteofAISafetyandGovernance(Beijing-AISI);BeijingKeyLaboratoryofSafeAIandSuperalignment;InternationalResearchCenterforAIEthicsandGovernance,InstituteofAutomation,ChineseAcademyofSciences.https://agile-index.ai/. AGILEIndexWebsite: https://agile-index.ai/(withsupportingdata&updatesacrosstime) ContributingInstitutes: CenterforLong-termArtificialIntelligence(CLAI)https://long-term-ai.center/BeijingInstituteofAISafetyandGovernance(Beijing-AISI)https://beijing-aisi.ac.cn/BeijingKeyLaboratoryofSafeAIandSuperalignmenthttps://beijing.safe-ai-and-superalignment.cn/InternationalResearchCenterforAIEthicsandGovernance,InstituteofAutomation,ChineseAcademyofScienceshttps://ai-ethics-and-governance.institute/ Fundinginformation: ThisresearchwasfinanciallysupportedbytheNationalScienceandTechnologyMajorProjectofChina(GrantNo.2022ZD0116202)andbytheBeijingMunicipalScience&TechnologyCommission. Contactinformation: Formoreinformationorwithanycommentsandfeedback,pleasecontactusat:contact@long-term-ai.cn. ReportVersion: Thisversion(v1.1.1)ofthereportwasofficiallypublishedonJuly23th,2025,succeedingthepre-releaseiteration(v1.0.0-pre)madeavailableonlineonJuly8th,2025. Copyrightinformation: Copyright©2025CenterforLong-termArtificialIntelligence-AllRightsReserved. ResearchTeam ChiefScientist YiZengisaProfessorfromInstituteofAutomation,ChineseAcademyofSciences,servingastheDirectoroftheInternationalResearchCenterforAIEthicsandGovernance,thePIoftheBrain-inspiredCognitiveAILab,andthe Director of Beijing Key Laboratory of Safe AI andSuperalignment.HeisthefoundingDeanofBeijingInstituteofAISafetyandGovernance(Beijing-AISI),thefoundingDirectoroftheCenterforLong-termAI,andleadstheAIforSDGsCooperationNetwork.HeisalsoamemberoftheUnitedNationsHigh-LevelAdvisoryBodyonAI,anexpertintheUNESCOAdHocExpertGroupontheEthicsofAI,anexpertintheWHOExpertGrouponEthicsandGovernanceofAIforHealth,aseniorfellowattheUNInstituteforDisarmamentResearch,aboardmemberoftheNationalGovernanceCommitteeofNextGenerationAIandaboardmemberoftheAIEthicsSub-CommitteeoftheNationalScienceandTechnologyEthicsCommitteeinChina,andaCo-ChairofWorldInternetConferenceSpecializedCommitteeonAI.Hisareasofinterestinclude Brain and Mind-inspired AI,AI Ethics,Safety andGovernance,andAIforGlobalSustainableDevelopment. YiZengDeanandProfessor Researchers BeijingInstituteofAISafetyandGovernanceBeijingKeyLaboratoryofSafeAIandSuperalignmentInstituteofAutomation,ChineseAcademyofSciencesBeijingKeyLaboratoryofSafeAIandSuperalignmentBeijingKeyLaboratoryofSafeAIandSuperalignmentBeijingInstituteofAISafetyandGovernanceInstituteofAutomation,ChineseAcademyofSciencesBeijingInstituteofAISafetyandGovernanceBeijingInstituteofAISafetyandGovernanceInstituteofAutomation,ChineseAcademyofSciencesCenterforLong-termAICenterforLong-termAI EnmengLuXiaoyangGuoCunqingHuangfuJiaweiXieYuChenZhengqiWangDongqiLiangGongceCaoJinWangZizheRuanXinGuanAmmarYounas What’sNewinAGILEIndex2025 DualExpansionofIndicators&CountryCoverage: Comparedtothe2024editionofAGILEIndex,thecountrycoverageinAGILEIndex2025hasexpandedfrom14to40,andthenumberofindicatorshasincreasedfrom39to43.Thisbroaderscopeprovidesastrongerfoundationforhorizontalcomparisonsandtrendanalysis. DataUpdate&DataSourceExpansion: Alldata from the previous edition have been thoroughly reviewed and updated where newerinformationwasavailablethroughMarch,2025.Morethan20additionaldatasourceshavealsobeenintegratedtoenhancecoverageandreliability. InclusionofFrontierAITechnologiesandResearchThemes: Newdata on cutting-edge AI technologies,such as generative AI,have been incorporated.Meanwhile,theliteratureanalysisnowcoversabroaderspectrumofAIsafetyandgovernancethemeswithrefinedkeywords,providingmorecomprehensiveoverviewofresearchadvancements. Year-on-YearComparisons: Forindicatorswithconsistentdatadefinitions,year-over-yearcomparisonshavebeenenhancedtobettercapturetrendsandhighlightkeychangesovertime. EnhancedMissingDataImputationStrategy: Inresponsetotheexpandedcountrycoverageandincreasingrequirementofdataqualityandscoring,thestrategyforhandlingmissingdatahasbeenfurtherrefined.Imputationnowcombineshistoricalperformancetrendsandcorrelationsbetweenindicatorstoensuregreaterrationalityandconsistency.PleaserefertotheMethodologysectionintheappendixfordetails. ExecutiveSummary ArtificialIntelligence(AI)representsbothanunprecedentedtechnologicalrevolutionandasystemicgovernancechallengeconfrontingtheglobal