人工智能、人權(quán)、民主和法治框架公約

CouncilofEuropeTreatySeries-No.225

CouncilofEuropeFrameworkConventiononArtificialIntelligenceandHumanRights,DemocracyandtheRuleofLaw

Vilnius,5.IX.2024

Preamble

ThememberStatesoftheCouncilofEuropeandtheothersignatorieshereto,

ConsideringthattheaimoftheCouncilofEuropeistoachievegreaterunitybetweenitsmembers,basedinparticularontherespectforhumanrights,democracyandtheruleoflaw;

Recognisingthevalueoffosteringco-operationbetweenthePartiestothisConventionandofextendingsuchco-operationtootherStatesthatsharethesamevalues;

Consciousoftheacceleratingdevelopmentsinscienceandtechnologyandtheprofoundchangesbroughtaboutthroughactivitieswithinthelifecycleofartificialintelligencesystems,whichhavethepotentialtopromotehumanprosperityaswellasindividualandsocietalwell-being,sustainabledevelopment,genderequalityandtheempowermentofallwomenandgirls,aswellasotherimportantgoalsandinterests,byenhancingprogressandinnovation;

Recognisingthatactivitieswithinthelifecycleofartificialintelligencesystemsmayofferunprecedentedopportunitiestoprotectandpromotehumanrights,democracyandtheruleoflaw;

Concernedthatcertainactivitieswithinthelifecycleofartificialintelligencesystemsmayunderminehumandignityandindividualautonomy,humanrights,democracyandtheruleoflaw;

Concernedabouttherisksofdiscriminationindigitalcontexts,particularlythoseinvolvingartificialintelligencesystems,andtheirpotentialeffectofcreatingoraggravatinginequalities,includingthoseexperiencedbywomenandindividualsinvulnerablesituations,regardingtheenjoymentoftheirhumanrightsandtheirfull,equalandeffectiveparticipationineconomic,social,culturalandpoliticalaffairs;

Concernedbythemisuseofartificialintelligencesystemsandopposingtheuseofsuchsystemsforrepressivepurposesinviolationofinternationalhumanrightslaw,includingthrougharbitraryorunlawfulsurveillanceandcensorshippracticesthaterodeprivacyandindividualautonomy;

Consciousofthefactthathumanrights,democracyandtheruleoflawareinherentlyinterwoven;

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

2

Convincedoftheneedtoestablish,asamatterofpriority,agloballyapplicablelegalframeworksettingoutcommongeneralprinciplesandrulesgoverningtheactivitieswithinthelifecycleofartificialintelligencesystemsthateffectivelypreservessharedvaluesandharnessesthebenefitsofartificialintelligenceforthepromotionofthesevaluesinamannerconducivetoresponsibleinnovation;

Recognisingtheneedtopromotedigitalliteracy,knowledgeabout,andtrustinthedesign,development,useanddecommissioningofartificialintelligencesystems;

RecognisingtheframeworkcharacterofthisConvention,whichmaybesupplementedbyfurtherinstrumentstoaddressspecificissuesrelatingtotheactivitieswithinthelifecycleofartificialintelligencesystems;

UnderliningthatthisConventionisintendedtoaddressspecificchallengeswhicharisethroughoutthelifecycleofartificialintelligencesystemsandencouragetheconsiderationofthewiderrisksandimpactsrelatedtothesetechnologiesincluding,butnotlimitedto,humanhealthandtheenvironment,andsocio-economicaspects,suchasemploymentandlabour;

Notingrelevanteffortstoadvanceinternationalunderstandingandco-operationonartificialintelligencebyotherinternationalandsupranationalorganisationsandfora;

Mindfulofapplicableinternationalhumanrightsinstruments,suchasthe1948UniversalDeclarationofHumanRights,the1950ConventionfortheProtectionofHumanRightsandFundamentalFreedoms(ETSNo.5),the1966InternationalCovenantonCivilandPoliticalRights,the1966InternationalCovenantonEconomic,SocialandCulturalRights,the1961EuropeanSocialCharter(ETSNo.35),aswellastheirrespectiveprotocols,andthe1996EuropeanSocialCharter(Revised)(ETSNo.163);

Mindfulalsoofthe1989UnitedNationsConventionontheRightsoftheChildandthe2006UnitedNationsConventionontheRightsofPersonswithDisabilities;

Mindfulalsooftheprivacyrightsofindividualsandtheprotectionofpersonaldata,asapplicableandconferred,forexample,bythe1981ConventionfortheProtectionofIndividualswithregardtoAutomaticProcessingofPersonalData(ETSNo.108)anditsprotocols;

AffirmingthecommitmentofthePartiestoprotectinghumanrights,democracyandtheruleof

law,andfosteringtrustworthinessofartificialintelligencesystemsthroughthisConvention,Haveagreedasfollows:

ChapterI-Generalprovisions

Article1-Objectandpurpose

1TheprovisionsofthisConventionaimtoensurethatactivitieswithinthelifecycleofartificialintelligencesystemsarefullyconsistentwithhumanrights,democracyandtheruleoflaw.

2EachPartyshalladoptormaintainappropriatelegislative,administrativeorothermeasurestogiveeffecttotheprovisionssetoutinthisConvention.Thesemeasuresshallbegraduatedanddifferentiatedasmaybenecessaryinviewoftheseverityandprobabilityoftheoccurrenceofadverseimpactsonhumanrights,democracyandtheruleoflawthroughoutthelifecycleofartificialintelligencesystems.Thismayincludespecificorhorizontalmeasuresthatapplyirrespectiveofthetypeoftechnologyused.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

3

3InordertoensureeffectiveimplementationofitsprovisionsbytheParties,thisConvention

establishesafollow-upmechanismandprovidesforinternationalco-operation.

Article2-Definitionofartificialintelligencesystems

ForthepurposesofthisConvention,“artificialintelligencesystem”meansamachine-basedsystemthat,forexplicitorimplicitobjectives,infers,fromtheinputitreceives,howtogenerateoutputssuchaspredictions,content,recommendationsordecisionsthatmayinfluencephysicalorvirtualenvironments.Differentartificialintelligencesystemsvaryintheirlevelsofautonomyandadaptivenessafterdeployment.

Article3-Scope

1ThescopeofthisConventioncoverstheactivitieswithinthelifecycleofartificialintelligencesystemsthathavethepotentialtointerferewithhumanrights,democracyandtheruleoflawasfollows:

aEachPartyshallapplythisConventiontotheactivitieswithinthelifecycleofartificialintelligencesystemsundertakenbypublicauthorities,orprivateactorsactingontheirbehalf.

bEachPartyshalladdressrisksandimpactsarisingfromactivitieswithinthelifecycleof

artificialintelligencesystemsbyprivateactorstotheextentnotcoveredinsubparagraphainamannerconformingwiththeobjectandpurposeofthisConvention.

EachPartyshallspecifyinadeclarationsubmittedtotheSecretaryGeneraloftheCouncilofEuropeatthetimeofsignature,orwhendepositingitsinstrumentofratification,acceptance,approvaloraccession,howitintendstoimplementthisobligation,eitherbyapplyingtheprinciplesandobligationssetforthinChaptersIItoVIofthisConventiontoactivitiesofprivateactorsorbytakingotherappropriatemeasurestofulfiltheobligationsetoutinthissubparagraph.Partiesmay,atanytimeandinthesamemanner,amendtheirdeclarations.

Whenimplementingtheobligationunderthissubparagraph,aPartymaynotderogatefromorlimittheapplicationofitsinternationalobligationsundertakentoprotecthumanrights,democracyandtheruleoflaw.

2APartyshallnotberequiredtoapplythisConventiontoactivitieswithinthelifecycleofartificialintelligencesystemsrelatedtotheprotectionofitsnationalsecurityinterests,withtheunderstandingthatsuchactivitiesareconductedinamannerconsistentwithapplicableinternationallaw,includinginternationalhumanrightslawobligations,andwithrespectforitsdemocraticinstitutionsandprocesses.

3WithoutprejudicetoArticle13andArticle25,paragraph2,thisConventionshallnotapplytoresearchanddevelopmentactivitiesregardingartificialintelligencesystemsnotyetmadeavailableforuse,unlesstestingorsimilaractivitiesareundertakeninsuchawaythattheyhavethepotentialtointerferewithhumanrights,democracyandtheruleoflaw.

4MattersrelatingtonationaldefencedonotfallwithinthescopeofthisConvention.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

4

ChapterII-Generalobligations

Article4-Protectionofhumanrights

EachPartyshalladoptormaintainmeasurestoensurethattheactivitieswithinthelifecycleofartificialintelligencesystemsareconsistentwithobligationstoprotecthumanrights,asenshrinedinapplicableinternationallawandinitsdomesticlaw.

Article5-Integrityofdemocraticprocessesandrespectfortheruleoflaw

1EachPartyshalladoptormaintainmeasuresthatseektoensurethatartificialintelligencesystemsarenotusedtounderminetheintegrity,independenceandeffectivenessofdemocraticinstitutionsandprocesses,includingtheprincipleoftheseparationofpowers,respectforjudicialindependenceandaccesstojustice.

2EachPartyshalladoptormaintainmeasuresthatseektoprotectitsdemocraticprocessesinthecontextofactivitieswithinthelifecycleofartificialintelligencesystems,includingindividuals’fairaccesstoandparticipationinpublicdebate,aswellastheirabilitytofreelyformopinions.

ChapterIII-PrinciplesrelatedtoactivitieswithinthelifecycleofartificialintelligencesystemsArticle6-Generalapproach

ThischaptersetsforthgeneralcommonprinciplesthateachPartyshallimplementinregardtoartificialintelligencesystemsinamannerappropriatetoitsdomesticlegalsystemandtheotherobligationsofthisConvention.

Article7-Humandignityandindividualautonomy

EachPartyshalladoptormaintainmeasurestorespecthumandignityandindividualautonomyinrelationtoactivitieswithinthelifecycleofartificialintelligencesystems.

Article8-Transparencyandoversight

EachPartyshalladoptormaintainmeasurestoensurethatadequatetransparencyandoversightrequirementstailoredtothespecificcontextsandrisksareinplaceinrespectofactivitieswithinthelifecycleofartificialintelligencesystems,includingwithregardtotheidentificationofcontentgeneratedbyartificialintelligencesystems.

Article9-Accountabilityandresponsibility

EachPartyshalladoptormaintainmeasurestoensureaccountabilityandresponsibilityforadverseimpactsonhumanrights,democracyandtheruleoflawresultingfromactivitieswithinthelifecycleofartificialintelligencesystems.

Article10-Equalityandnon-discrimination

1EachPartyshalladoptormaintainmeasureswithaviewtoensuringthatactivitieswithinthelifecycleofartificialintelligencesystemsrespectequality,includinggenderequality,andtheprohibitionofdiscrimination,asprovidedunderapplicableinternationalanddomesticlaw.

2EachPartyundertakestoadoptormaintainmeasuresaimedatovercominginequalitiestoachievefair,justandequitableoutcomes,inlinewithitsapplicabledomesticandinternationalhumanrightsobligations,inrelationtoactivitieswithinthelifecycleofartificialintelligencesystems.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

5

Article11-Privacyandpersonaldataprotection

EachPartyshalladoptormaintainmeasurestoensurethat,withregardtoactivitieswithinthelifecycleofartificialintelligencesystems:

aprivacyrightsofindividualsandtheirpersonaldataareprotected,includingthroughapplicabledomesticandinternationallaws,standardsandframeworks;and

beffectiveguaranteesandsafeguardshavebeenputinplaceforindividuals,inaccordancewithapplicabledomesticandinternationallegalobligations.

Article12-Reliability

EachPartyshalltake,asappropriate,measurestopromotethereliabilityofartificialintelligencesystemsandtrustintheiroutputs,whichcouldincluderequirementsrelatedtoadequatequalityandsecuritythroughoutthelifecycleofartificialintelligencesystems.

Article13-Safeinnovation

Withaviewtofosteringinnovationwhileavoidingadverseimpactsonhumanrights,democracyandtheruleoflaw,eachPartyiscalledupontoenable,asappropriate,theestablishmentofcontrolledenvironmentsfordeveloping,experimentingandtestingartificialintelligencesystemsunderthesupervisionofitscompetentauthorities.

ChapterIV-Remedies

Article14-Remedies

1EachPartyshall,totheextentremediesarerequiredbyitsinternationalobligationsandconsistentwithitsdomesticlegalsystem,adoptormaintainmeasurestoensuretheavailabilityofaccessibleandeffectiveremediesforviolationsofhumanrightsresultingfromtheactivitieswithinthelifecycleofartificialintelligencesystems.

2Withtheaimofsupportingparagraph1above,eachPartyshalladoptormaintainmeasuresincluding:

ameasurestoensurethatrelevantinformationregardingartificialintelligencesystemswhichhavethepotentialtosignificantlyaffecthumanrightsandtheirrelevantusageisdocumented,providedtobodiesauthorisedtoaccessthatinformationand,whereappropriateandapplicable,madeavailableorcommunicatedtoaffectedpersons;

bmeasurestoensurethattheinformationreferredtoinsubparagraphaissufficientfortheaffectedpersonstocontestthedecision(s)madeorsubstantiallyinformedbytheuseofthesystem,and,whererelevantandappropriate,theuseofthesystemitself;and

caneffectivepossibilityforpersonsconcernedtolodgeacomplainttocompetentauthorities.

Article15-Proceduralsafeguards

1EachPartyshallensurethat,whereanartificialintelligencesystemsignificantlyimpactsupontheenjoymentofhumanrights,effectiveproceduralguarantees,safeguardsandrights,inaccordancewiththeapplicableinternationalanddomesticlaw,areavailabletopersonsaffectedthereby.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

6

2EachPartyshallseektoensurethat,asappropriateforthecontext,personsinteractingwithartificialintelligencesystemsarenotifiedthattheyareinteractingwithsuchsystemsratherthanwithahuman.

ChapterV-AssessmentandmitigationofrisksandadverseimpactsArticle16-Riskandimpactmanagementframework

1EachPartyshall,takingintoaccounttheprinciplessetforthinChapterIII,adoptormaintainmeasuresfortheidentification,assessment,preventionandmitigationofrisksposedbyartificialintelligencesystemsbyconsideringactualandpotentialimpactstohumanrights,democracyandtheruleoflaw.

2Suchmeasuresshallbegraduatedanddifferentiated,asappropriate,and:

atakedueaccountofthecontextandintendeduseofartificialintelligencesystems,in

particularasconcernsriskstohumanrights,democracy,andtheruleoflaw;btakedueaccountoftheseverityandprobabilityofpotentialimpacts;

cconsider,whereappropriate,theperspectivesofrelevantstakeholders,inparticularpersonswhoserightsmaybeimpacted;

dapplyiterativelythroughouttheactivitieswithinthelifecycleoftheartificialintelligencesystem;

eincludemonitoringforrisksandadverseimpactstohumanrights,democracy,andtheruleoflaw;

fincludedocumentationofrisks,actualandpotentialimpacts,andtheriskmanagement

approach;and

grequire,whereappropriate,testingofartificialintelligencesystemsbeforemakingthemavailableforfirstuseandwhentheyaresignificantlymodified.

3EachPartyshalladoptormaintainmeasuresthatseektoensurethatadverseimpactsofartificialintelligencesystemstohumanrights,democracy,andtheruleoflawareadequatelyaddressed.Suchadverseimpactsandmeasurestoaddressthemshouldbedocumentedandinformtherelevantriskmanagementmeasuresdescribedinparagraph2.

4EachPartyshallassesstheneedforamoratoriumorbanorotherappropriatemeasuresinrespectofcertainusesofartificialintelligencesystemswhereitconsiderssuchusesincompatiblewiththerespectforhumanrights,thefunctioningofdemocracyortheruleoflaw.

ChapterVI-ImplementationoftheConventionArticle17-Non-discrimination

TheimplementationoftheprovisionsofthisConventionbythePartiesshallbesecuredwithoutdiscriminationonanyground,inaccordancewiththeirinternationalhumanrightsobligations.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

7

Article18-Rightsofpersonswithdisabilitiesandofchildren

EachPartyshall,inaccordancewithitsdomesticlawandapplicableinternationalobligations,takedueaccountofanyspecificneedsandvulnerabilitiesinrelationtorespectfortherightsofpersonswithdisabilitiesandofchildren.

Article19-Publicconsultation

EachPartyshallseektoensurethatimportantquestionsraisedinrelationtoartificialintelligencesystemsare,asappropriate,dulyconsideredthroughpublicdiscussionandmultistakeholderconsultationinthelightofsocial,economic,legal,ethical,environmentalandotherrelevantimplications.

Article20-Digitalliteracyandskills

EachPartyshallencourageandpromoteadequatedigitalliteracyanddigitalskillsforallsegmentsofthepopulation,includingspecificexpertskillsforthoseresponsiblefortheidentification,assessment,preventionandmitigationofrisksposedbyartificialintelligencesystems.

Article21-Safeguardforexistinghumanrights

NothinginthisConventionshallbeconstruedaslimiting,derogatingfromorotherwiseaffectingthehumanrightsorotherrelatedlegalrightsandobligationswhichmaybeguaranteedundertherelevantlawsofaPartyoranyotherrelevantinternationalagreementtowhichitisparty.

Article22-Widerprotection

NoneoftheprovisionsofthisConventionshallbeinterpretedaslimitingorotherwiseaffectingthepossibilityforaPartytograntawidermeasureofprotectionthanisstipulatedinthisConvention.

ChapterVII-Follow-upmechanismandco-operationArticle23-ConferenceoftheParties

1TheConferenceofthePartiesshallbecomposedofrepresentativesofthePartiestothisConvention.

2ThePartiesshallconsultperiodicallywithaviewto:

afacilitatingtheeffectiveapplicationandimplementationofthisConvention,includingtheidentificationofanyproblemsandtheeffectsofanyreservationmadeinpursuanceof

Article34,paragraph1,oranydeclarationmadeunderthisConvention;

bconsideringthepossiblesupplementationtooramendmentofthisConvention;

cconsideringmattersandmakingspecificrecommendationsconcerningtheinterpretationandapplicationofthisConvention;

dfacilitatingtheexchangeofinformationonsignificantlegal,policyortechnologicaldevelopmentsofrelevance,includinginpursuitoftheobjectivesdefinedinArticle25,fortheimplementationofthisConvention;

efacilitating,wherenecessary,thefriendlysettlementofdisputesrelatedtotheapplicationofthisConvention;and

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

8

ffacilitatingco-operationwithrelevantstakeholdersconcerningpertinentaspectsofthe

implementationofthisConvention,includingthroughpublichearingswhereappropriate.

3TheConferenceofthePartiesshallbeconvenedbytheSecretaryGeneraloftheCouncilofEuropewhenevernecessaryand,inanycase,whenamajorityofthePartiesortheCommitteeofMinistersrequestsitsconvocation.

4TheConferenceofthePartiesshalladoptitsownrulesofprocedurebyconsensuswithintwelvemonthsoftheentryintoforceofthisConvention.

5ThePartiesshallbeassistedbytheSecretariatoftheCouncilofEuropeincarryingouttheirfunctionspursuanttothisarticle.

6TheConferenceofthePartiesmayproposetotheCommitteeofMinistersappropriatewaystoengagerelevantexpertiseinsupportoftheeffectiveimplementationofthisConvention.

7AnyPartywhichisnotamemberoftheCouncilofEuropeshallcontributetothefundingoftheactivitiesoftheConferenceoftheParties.Thecontributionofanon-memberoftheCouncilofEuropeshallbeestablishedjointlybytheCommitteeofMinistersandthatnon-member.

8TheConferenceofthePartiesmaydecidetorestricttheparticipationinitsworkofaPartythathasceasedtobeamemberoftheCouncilofEuropeunderArticle8oftheStatuteoftheCouncilofEurope(ETSNo.1)foraseriousviolationofArticle3oftheStatute.Similarly,measurescanbetakeninrespectofanyPartythatisnotamemberStateoftheCouncilofEuropebyadecisionoftheCommitteeofMinisterstoceaseitsrelationswiththatStateongroundssimilartothosementionedinArticle3oftheStatute.

Article24-Reportingobligation

1EachPartyshallprovideareporttotheConferenceofthePartieswithinthefirsttwoyearsafterbecomingaParty,andthenperiodicallythereafterwithdetailsoftheactivitiesundertakentogiveeffecttoArticle3,paragraph1,sub-paragraphsaandb.

2TheConferenceofthePartiesshalldeterminetheformatandtheprocessforthereportinaccordancewithitsrulesofprocedure.

Article25-Internationalco-operation

1ThePartiesshallco-operateintherealisationofthepurposeofthisConvention.Partiesarefurtherencouraged,asappropriate,toassistStatesthatarenotPartiestothisConventioninactingconsistentlywiththetermsofthisConventionandbecomingaPartytoit.

2ThePartiesshall,asappropriate,exchangerelevantandusefulinformationbetweenthemselvesconcerningaspectsrelatedtoartificialintelligencewhichmayhavesignificantpositiveornegativeeffectsontheenjoymentofhumanrights,thefunctioningofdemocracyandtheobservanceoftheruleoflaw,includingrisksandeffectsthathaveariseninresearchcontextsandinrelationtotheprivatesector.Partiesareencouragedtoinvolve,asappropriate,relevantstakeholdersandStatesthatarenotPartiestothisConventioninsuchexchangesofinformation.

CETS225-ArtificialIntelligence,5.IX.2024

_____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

9

3ThePartiesareencouragedtostrengthenco-operation,includingwithrelevantstakeholderswhereappropriate,topreventandmitigaterisksandadverseimpactsonhumanrights,democracyandtheruleoflawinthecontextofactivitieswithinthelifecycleofartificialintelligencesystems.

Article26-Effectiveoversightmechanisms

1EachPartyshallestablishordesignateoneormoreeffectivemechanismstooverseecompliancewiththeobligationsinthisConvention.

2EachPartyshallensurethatsuchmechanismsexercisetheirdutiesindependentlyandimpartiallyandthattheyhavethenecessarypowers,expertiseandresourcestoeffectivelyfulfiltheirtasksofoverseeingcompliancewiththeobligationsinthisConvention,asgiveneffectbytheParties.

3IfaPartyhasprovidedformorethanonesuchmechanism,itshalltakemeasures,wherepracticable,tofacilitateeffectivecooperationamongthem.

4IfaPartyhasprovidedformechanismsdifferentfromexistinghumanrightsstructures,itshalltakemeasures,wherepracticable,topromoteeffectivecooperationbetweenthemechanismsreferredtoinparagraph1andthoseexistingdomestichumanrightsstructures.

ChapterVIII-Finalclauses

Article27-EffectsoftheConvention

1IftwoormorePartieshavealreadyconcluded