2025保障前沿人工智能（AI）的治理路径研究报告（英文）.pdfVIP

ResearchReport

IANMITCH,MATTHEWJ.MALONE,KARENSCHWINDT,GREGORYSMITH,WESLEYHURD,

HENRYALEXANDERBRADLEY,JAMESGIMBI

Governance

Approachesto

SecuringFrontierAI

ThetitleofthispublicationwasupdatedinOctober2025.

Formoreinformationonthispublication,visit/t/RRA4159-1.

AboutRAND

RANDisaresearchorganizationthatdevelopssolutionstopublicpolicychallengestohelpmakecommunitiesthroughouttheworld

saferandmoresecure,healthierandmoreprosperous.RANDisnonprofit,nonpartisan,andcommittedtothepublicinterest.Tolearn

moreaboutRAND,visit.

ResearchIntegrity

Ourmissiontohelpimprovepolicyanddecisionmakingthroughresearchandanalysisisenabledthroughourcorevaluesofquality

andobjectivityandourunwaveringcommitmenttothehighestlevelofintegrityandethicalbehavior.Tohelpensureourresearch

andanalysisarerigorous,objective,andnonpartisan,wesubjectourresearchpublicationstoarobustandexactingquality-assurance

process;avoidboththeappearanceandrealityoffinancialandotherconflictsofinterestthroughstafftraining,projectscreening,

andapolicyofmandatorydisclosure;andpursuetransparencyinourresearchengagementsthroughourcommitmenttotheopen

publicationofourresearchfindingsandrecommendations,disclosureofthesourceoffundingofpublishedresearch,andpoliciesto

ensureintellectualindependence.Formoreinformation,visit/about/research-integrity.

RAND’spublicationsdonotnecessarilyreflecttheopinionsofitsresearchclientsandsponsors.

PublishedbytheRANDCorporation,SantaMonica,Calif.

©2025RANDCorporation

isaregisteredtrademark.

Cover:AdobeStock/Artofinnovation.

LimitedPrintand