The OECD Artificial Intelligence (AI) Principles

The OECD AI Principles promote use of AI that is innovative and trustworthy and that respects human rights and democratic values. Adopted in May 2019, ... OECD.orgGoingDigitalToolkit|enExperts&blogAIWonkblogOECDNetworkofExpertsonAIOECDParliamentaryGrouponAIGlobalPartnershiponAIAIpolicyeventsSignuptoourmailinglistAIPrinciplesOECDAIPrinciplesoverviewPolicyareasPolicyareasoverviewWork,Innovation,Productivity&SkillsprogrammeTrends&dataOECDmetrics&methodsCountries&initiativesNationalstrategies&policiesStakeholderinitiativesAboutAboutOECD.AIResourcesonAIPartnersFAQVideosContactusSearchHomeOECDAIPrinciplesoverviewOECDAIPrinciplesoverviewTheOECDAIPrinciplespromoteuseofAIthatisinnovativeandtrustworthyandthatrespectshumanrightsanddemocraticvalues.AdoptedinMay2019,theysetstandardsforAIthatarepracticalandflexibleenoughtostandthetestoftime.Values-basedprinciplesRecommendationsforpolicymakersTheOECDAIPrinciplesfocusonhowgovernmentsandotheractorscanshapeahuman-centricapproachtotrustworthyAI.AsanOECDlegalinstrument,theprinciplesrepresentacommonaspirationforitsadheringcountries.OfficialtextGovernmentsthathavecommittedtotheAIPrinciplesEnlargeAIterms&conceptsAIsystemAIsystemlifecycleAIknowledgeAIactorsStakeholdersTrustworthyAIAnAIsystemisamachine-basedsystemthatiscapableofinfluencingtheenvironmentbyproducinganoutput(predictions,recommendationsordecisions)foragivensetofobjectives.Itusesmachineand/orhuman-baseddataandinputsto(i)perceiverealand/orvirtualenvironments;(ii)abstracttheseperceptionsintomodelsthroughanalysisinanautomatedmanner(e.g.,withmachinelearning),ormanually;and(iii)usemodelinferencetoformulateoptionsforoutcomes.AIsystemsaredesignedtooperatewithvaryinglevelsofautonomy.EnlargeTheOECD’sworkonArtificialIntelligenceandrationalefordevelopingtheOECDRecommendationonArtificialIntelligenceAIisageneral-purposetechnologythathasthepotentialtoimprovethewelfareandwell-beingofpeople,tocontributetopositivesustainableglobaleconomicactivity,toincreaseinnovationandproductivity,andtohelprespondtokeyglobalchallenges.Itisdeployedinmanysectorsrangingfromproduction,financeandtransporttohealthcareandsecurity.Alongsidebenefits,AIalsoraiseschallengesforoursocietiesandeconomies,notablyregardingeconomicshiftsandinequalities,competition,transitionsinthelabourmarket,andimplicationsfordemocracyandhumanrights.TheOECDhasundertakenempiricalandpolicyactivitiesonAIinsupportofthepolicydebateoverthepasttwoyears,startingwithaTechnologyForesightForumonAIin2016andaninternationalconferenceonAI:IntelligentMachines,SmartPoliciesin2017.TheOrganisationalsoconductedanalyticalandmeasurementworkthatprovidesanoverviewoftheAItechnicallandscape,mapseconomicandsocialimpactsofAItechnologiesandtheirapplications,identifiesmajorpolicyconsiderations,anddescribesAIinitiativesfromgovernmentsandotherstakeholdersatnationalandinternationallevels.ThisworkhasdemonstratedtheneedtoshapeastablepolicyenvironmentattheinternationalleveltofostertrustinandadoptionofAIinsociety.Againstthisbackground,theOECDCommitteeonDigitalEconomyPolicy(CDEP)agreedtodevelopadraftCouncilRecommendationtopromoteahuman-centricapproachtotrustworthyAI,thatfostersresearch,preserveseconomicincentivestoinnovate,andappliestoallstakeholders.AninclusiveandparticipatoryprocessfordevelopingtheRecommendationThedevelopmentoftheRecommendationwasparticipatoryinnature,incorporatinginputfromabroadrangeofsourcesthroughouttheprocess.InMay2018,theCDEPagreedtoformanexpertgrouptoscopeprinciplestofostertrustinandadoptionofAI,withaviewtodevelopingadraftCouncilRecommendationinthecourseof2019.TheAIGroupofexpertsattheOECD(AIGO)wassubsequentlyestablished,comprisingover50expertsfromdifferentdisciplinesanddifferentsectors(government,industry,civilsociety,tradeunions,thetechnicalcommunityandacademia)-seethefulllist.BetweenSeptember2018andFebruary2019thegroupheldfourmeetings:inParis,France,inSeptemberandNovember2018,inCambridge,MA,UnitedStates,attheMassachusettsInstituteofTechnology(MIT)inJanuary2019,backtobackwiththeMITAIPolicyCongress,andfinallyinDubai,UnitedArabEmirates,attheWorldGovernmentSummitinFebruary2019.Theworkbenefitedfromthediligence,engagementandsubstantivecontributionsoftheexpertsparticipatinginAIGO,aswellasfromtheirmulti-stakeholderandmultidisciplinarybackgrounds.DrawingonthefinaloutputdocumentoftheAIGO,adraftRecommendationwasdevelopedintheCDEPandwiththeconsultationofotherrelevantOECDbodies.TheCDEPapprovedafinaldraftRecommendationandagreedtotransmitittotheOECDCouncilforadoptioninaspecialmeetingon14-15March2019.TheOECDCounciladoptedtheRecommendationatitsmeetingatMinisteriallevelon22-23May2019.MentionedresourcesGovernmentsthathavecommittedtotheAIPrinciples