AIgovernanceandstandards

Policypaper

Executivesummary 2

1.Whyweneedstandardsandwhattheyhelpusachieve 2

Whatarestandards? 2

HowarestandardsandothersoftlawmechanismsusedinrelationtoAItechnologies? 3

2.Thecurrentstandardslandscape 4

GovernmentactivitiesinAIstandardsdevelopment 4

WhatarethebenefitsofstandardsforAI,whatarethechallengesandhowcanwe

overcomethem? 5

Potentialforoverlaps,duplicationsanddivergencesinAIstandards 6

3.Recommendations 8

July2025|AIgovernanceandstandards|1

July2025|AIgovernanceandstandards|2

Executivesummary

Thepaperrespondstothegrowingriskoffragmentationinglobalgovernanceofartificial

intelligence(AI).AscountriesandregionsdeveloptheirownAIlaws,policiesandregulations,

andevenstandards,divergentapproachesareemerging.Thepaperhighlightstheimportantroleofinternationalandmarket-drivenstandardsinsupportingabroadrangeofgovernment

approachestoAIoversightanddiscusseshowstandardscanhelpbridgelegaldifferencesacrossjurisdictions.

Standardscanprovideconsistent,practicalsolutionsandguidancetocomplywithlaws,policies

andregulation.Whengovernmentsreferencestandardsasthemeanstoimplementnon-technicalorhigh-levelperformancerequirements,theyavoidwritingtechnicalrequirementsthatmay

introduceunnecessarycoststomakeproducts/servicesavailableintheirjurisdiction.Nationalorregionaltechnicalrequirementsintroducedthroughregulationorstandardscreatecomplexityforbusinessesofallsizes,increasecompliancecosts,limitnationalproductivitygainsbyimpedingAIadoptionandriskslowingcross-bordercollaborationandinnovation.

ToensureeffectiveandinteroperableAIgovernance,greateradoptionofstandardsisessential.Businessesandgovernmentscanalsobringcriticalexpertiseandoperationalinsighttothe

standardsdevelopmentprocessitself.Promotingtheuseofmarket-drivenstandardscanreduceduplication,improveregulatorycoherenceandsupportpolicyobjectives.

1.Whyweneedstandardsandwhattheyhelpusachieve

FromcompaniesdevelopingalgorithmstothosedeployingAIservicesorsystemsfortheirend-users,eachparticipantintheAIsupplychainneedsclear,consistentguidelines.International

standardsareavitaltoolforestablishingthesesharedexpectations.Theyserveasafoundationforfosteringinteroperability,providingthemeansforregulatoryalignmentandfacilitatingthe

globaldisseminationofAIinnovation.

Standardsbodiesmaintainthestandardstheyproduceandregularlydetermineifeachstandardshouldberevised,confirmedorwithdrawn.ThisabilitytoevolvealongsideAItechnologiesisan

importantfeatureofstandards(incomparisontoothermechanisms)thatkeepthemrelevantandeffectiveovertime.

Whatarestandards?

Standardsaredocumentsspecifyingrequirements,guidelines,orcharacteristicsofaproduct,service,process,orsystem.

Theyaredevelopedinrules-based,voluntary,multistakeholderorganisationsthatcanbe

horizontalorsectoralandcanbenational,regionalorinternational.Examplesofprominent

informationtechnologyinternationalandmarket-drivenstandardsbodiesincludetheInternationalOrganisationforStandardisation(ISO),theInternationalElectrotechnicalCommission(IEC),the

InternationalTelecommunicationsUnion(ITU),theInstituteofElectricalandElectronicsEngineers(IEEE),theInternetEngineeringTaskForce(IETF)andtheEuropeanTelecommunicationsStandardsInstitute(ETSI).

Standardsdevelopmentorganisationstendtohavematuregovernancesystems,particularlyduetoprocessrequirements1thattakethetimeneededtoelicitcontributionsfromabroadsetof

1

WorldTradeOrganisation(WTO),PrinciplesfortheDevelopmentofInternationalStandards,Guidesand

Recommendations(2024).

July2025|AIgovernanceandstandards|3

stakeholdersandestablishconsensusamongthem.Theprocessisintendedtoproducehigh-

qualityresultsthatreflectthebestavailabletechnicalsolutionsorguidancetobuildpublictrust

andlegitimacyespeciallyinsectorslikehealth,justiceandeducation.Regulationcanbenefitfromthisdeliberativeprocessbycitingstandardstoframearegulatoryapproachorsetmorespecificrequirementstocarryouttheregulation.

HowarestandardsandothersoftlawmechanismsusedinrelationtoAItechnologies?

Thereareavarietyofsoftlawmechanismsavailablethatcanbeusedtogoverninformation

technology:standards,open-sourcesoftware(OSS),high-levelexpertgroupfindings,requirementsmandatedthroughthesupplychain(i.e.requirementsimposedthroughprocurement),codesof

conductandguidelines.

AIsystemsandsolutionsareoftenbuiltfrompartsproducedbydifferentactors,oftenindifferentjurisdictions.Standardscontributetothevitalobjectiveofassuringresponsible,safe,secureandinteroperableAIsystemsandsolutionsbyfosteringtechnicalconsistencyandregulatory

alignment.Thisharmonisationisessentialformaintainingconsistencyinthesepracticesand

complianceacrossglobalmarkets.InthefieldofAI,somestandardsaredevelopedwiththegoalofharmonisingfoundationalconceptsortopromoteresponsibleAImanagementpractices.

Standardscanalsoprovidethemeanstoaddressbroadprinciples,includingprinciplesthataredefinedinlawsandregulation.Otherstandardsarebeingdevelopedtomanagesafetyand

securityriskstoprotecttheinformationmanagedonthosesystemsaswellasthepersonsandorganisationsinvolved.RiskmanagementandAIgovernancearetwoareaswherestandardsprovideconsistencyofconceptsandapproachtoriskmanagementandsystemassurance,asdiscussedinthe

casestudiesbelow.

Othersoftlawmechanismsalsoplayvaluablerolesandaresometimesconflatedwithstandards.Ininformationtechnology(IT)andoperationaltechnology(OT),technicalinteroperabilityisoftensupportedbytheuseofOSSimplementationwhichservesasamethodologyforcollaborative

softwaredevelopment.OSSiswidelyusedtodefineapplicationprogramminginterfaces(APIs)andprotocols.

Technology-specificprinciples,codesofconductandguidelines(e.g.secureandsafesoftware

engineering)canbedevelopedtooutlinethenormsofbehaviourand/orbestpracticeexpected

fromacertaingroupofactorsorexperts,includingtodefinehowtocomplywithlawsand

regulations.Forexample,theOrganizationforEconomicCooperationandDevelopment(OECD)

adoptedgroundbreakingAIPrinciplesin20192whichhavebeenformallyadheredtobytheOECD’s38membercountries,theEuropeanUnionandnineothercountries.Anotherexampleisthe

HiroshimaCodeofConductforOrganizationsDevelopingAdvancedAISystems3adoptedbytheG7in2023,onwhichcompaniescanvoluntarilycontributetoanOECDreportingframework4ontheiractionstoadheretothecode.

ManyorganisationsstruggletoadoptAIstandardsduetoalackofawareness,technical

expertise,orregulatoryclarity.Regardingthesecondfactor(technicalexpertise),whilethere

isagrowingbodyofguidelinesandframeworks,thereisstillaneedfordetailedpractical

informationtoguidethecomprehensiveandcoherentimplementationofstandardswhichservealsotoguideorganisationsonhowtocomplywithgrowingregulatorydevelopments.These

2

OrganisationforEconomicCo-operationandDevelopment(OECD),OECDAIPrinciples(2019,updated2024).

3

G7,HiroshimaProcessInternationalCodeofConductforOrganizationsDevelopingAdvancedAISystems(2023).

4

G7,HiroshimaProcessReportingFramework(2025).

July2025|AIgovernanceandstandards|4

kindsofguideswithpracticalinformationarebestdevelopedthroughcommunity-driven

initiativesthatcandevelopresources(e.g.specificusecaseguidance,open-sourcetools)

outsideofthemoreformalprocessusedtodeterminewhereandhowstakeholderscanagreetorequirementsinstandards.

Internationalstandardscansignificantlylowerbarrierstoentryforsmall-andmedium-sizedenterprises(SMEs)byprovidingscalablesolutionsforbothcustomerassuranceand

regulatorycompliance.

2.Thecurrentstandardslandscape

Worldwide,therearehundredsofprivateorganisationsdevelopingITstandards.Amongthose,

therearearelativelysmallnumberoforganisationsdevelopingAIstandards.Thesespecialised

bodiesarepivotalinaddressingtheuniquechallengesposedbyAItechnologies,includingdigitalcontenttransparency,security,fairnessandaccountability.Mostoftheseorganisationshave

manystandardsprojectsunderdevelopment.

TherapidpaceofAIdevelopmentdemandsagileandadaptablestandardsthatcankeepupwithevolvingtechnologies.Standardsdevelopmentmaybeanticipatoryorreactionary(orsomewhereinbetween)withrespecttoproductsandservicesenteringthemarketplace.OneofthestrengthsoftheAIstandards-setting‘system’hasbeenitsabilitytoactuponglobalrecognitionoftheneedforinteroperablestandards.

Infact,AI-specificstandardsstartedbefore2020,andthedevelopmentofAIstandardsis

progressingrapidlyacrossarangeofnational,regional,andinternationalbodies.Manyinitiativesemphasisetransparency,fairness,safetyandaccountabilityinAIsystems,helpingtoaligneffortstowardtrustworthyAI.

Globally,variousorganisations,includingtheISO,IEC,IEEE,ITU,EuropeanCommitteefor

Standardization(CEN)andtheEuropeanCommitteeforElectrotechnicalStandardization

(CENELEC),ETSIandothersmallerindustry-ledconsortia,e.g.CoalitionforContentProvenanceandAuthenticity(C2PA),areactivelyworkingonstandardstoguidetheresponsibledevelopmentanddeploymentofAI.

InEurope,CENandtheCENELECareplayingacrucialroleinaligningAIstandardswiththeEUAIAct.InMay2023,theEuropeanCommissiontaskedtheCENandCENELECtodevelopstandardsfortheEUAIAct’s5high-riskprovisions.TheEUAIActrequireshigh-riskAIsystemprovidersto

implementqualityandriskmanagementsystemsevenafteraproductisplacedonthemarket.HarmonisedEuropeanstandardswillprovideapathtomeettheserequirements.

Additionally,intergovernmentalinitiativessuchasthoseledbytheOECD,UNESCOortheCouncilofEuropearecontributingtotheconversationbyestablishingterminologyandprinciplesthat

contributetothedevelopmentoftechnicalandregulatorystandards.Industry-drivenstandardsplayavitalroleintranslatinghigh-levelprinciplesandinitiativesintopractical,interoperable

solutions,providingguidanceonAIgovernance,riskmanagementandsecurityconsiderations.

GovernmentactivitiesinAIstandardsdevelopment

GovernmentsareengagedindevelopingAIstandardstoguidetheresponsibleuseofAI.IntheUnitedStates,theNationalInstituteofStandardsandTechnology(NIST)hasbeenactively

5

EuropeanUnion,EUArtificialIntelligenceAct(2024).

July2025|AIgovernanceandstandards|5

workingalsoindevelopingstandardsinalignmentwiththeExecutiveOrders(fromtheWhite

House)aswellasthelawsandregulationsinprogressonthematter.ManyofthosestandardsareactivelyusedinCanadaandLatinAmerica.Atthesametime,theEUisworkingonthefirstCode

ofPractice6forprovidersofgeneral-purposeAImodels,includingthoseconsideredtopose

systemicrisks,inanticipationoftheEUAIActcomingintoeffectinAugust2025forthesemodels.

TheCodeofPracticeisintendedtoserveasabridgeaheadofthedevelopmentandavailabilityofformalstandards.Regionalorsectoralstandardsshouldremainfullycompatiblewithand

whereverpossibleidenticaltoexistinginternationalstandardssuchasISO/IECstandardstopreventmarketfragmentation.

WhatarethebenefitsofstandardsforAI,whatarethechallengesandhowcanweovercomethem?

Internationalandmarket-drivenstandardscanplayakeyroleinfosteringgloballyinteroperableAIgovernanceanddriveinteroperability.Thisisimportantformakingiteasierfororganisationsandcompaniestocollaborateacrossbordersbyprovidingwaystoconformwithregulatoryand/or

customerrequirementsatagloballevel,accessthebestproductsandtools,andenablethe

benefitsofAItobespreadasbroadlyaspossible.Duplicativeandpotentiallyconflictingstandardsandcomplianceschemes,however,raisethecostsofdoingbusinessinanincreasinglyglobalisedworld,underminingthispotentialforinteroperability.

Thegrowingdevelopmentofglobal,regionalandnationalAIpolicies,lawsandregulationsrisks

creatingdivergentgovernanceapproachesandcreatingacomplexregulatorylandscapewhichhindersthepotentialtospreadthetechnologiesacrossborders.Inaddition,therecanbe

inconsistenciesbetweenpoliciesandtechnicalstandards,giventhatstandardsoftenemerge

fromtheneedsoftechnologydevelopersanddeployersandnotfromregulatoryneeds.However,moreoften,standardsprovideacommonapproach,evenwherelegalandregulatoryapproachesdifferbetweencountriesandregions.Inaddition,theyenhancetrustamongconsumersand

businessesinAItechnologies.

Referencingstandardsinregulationscanexplainhowtomeettherequirementsorfacilitatethe

implementationofaregulation,buttheycannotextendregulation.Saidanotherway,theyarenotasubstitutefortheroleofgovernments.Onewaytoaddresstheseissuesisthereforefor

policymakerstoconsiderreferringtomarket-drivenstandardswhendesigninglawsanddraftingregulationsorallowforconformancewithastandardtobeconsideredsufficientformeeting

regulatoryrequirements.Standardscanthussupportregulation,andpolicymakerscanbenefitfromtakingthemintoaccount.

Relatedly,theGlobalDigitalCompact(GDC)7,adoptedinSeptember2024bytheUNGeneralAssembly,calledonstandardsdevelopingorganisations“tocollaboratetopromotethe

developmentandadoptionofinteroperableAIstandardsthatupholdsafety,reliability,

sustainabilityandhumanrights”.SuchcollaborationandcoordinationaroundAIstandardsisimportanttopromoteinteroperabilityandsupportpolicymakers’useofstandards.

Inadditiontoreferencingstandards(specificallyorgenerally)inlawsandregulations,itis

importantforstandardstobeincorporatedintoprocurementprocessesovergovernment-uniquestandardsortechnicalrequirements.Thisisatoolfordrivinginteroperabilityandavoiding

fragmentation,asitcanhelpguidesmallercompanies,oftennotinvolvedinorawareofthe

6

EuropeanCommission,General-PurposeAICodeofPractice(effectiveAugust2025).

7UN,GlobalDigitalCompact,(2024).

July2025|AIgovernanceandstandards|6

standardsdevelopmentprocess,touseacommonapproach.Astrustmechanisms,governancestandardscanalsofacilitatecommercialcontractinganddemonstrateconformitytoregulations.

Potentialforoverlaps,duplicationsanddivergencesinAIstandards

Thereareseveralfactorsthatmayresultinoverlaps,duplicationsordivergencesinAIstandards:

1.Overlapsandduplications:

?Regulatoryfragmentation:WhileinternationalbodieslikeISOandIECaredeveloping

foundational/horizontalAIstandardsthatalignwithregulationatahighlevelwhileremaining

country/regionagnostic(e.g.ISO/IEC420018,238949and4200510),theEuropeanCommission

(EC)hasrequestedthatCEN-CENELECdevelopstandardstocarryoutrequirementsoftheEUAIAct,potentiallyfragmentingmarketsand/orcreatingoverlappingbutslightlydifferingtechnicalrequirements.InthecaseofNISTAIdocuments,theycanalsooverlapbecausetheycanarise

fromUSExecutiveOrdersand/orUSlawsandregulations.

?Proliferationofstandards:StandardsorganisationsarecompelledtostartnewprojectstoaddressnewtrendsinAI,whenlimitedornochangestoexistingstandardsaresufficient.

?Policy-drivenstandardsetting:Someparticipantsattempttoaddresstheirspecificpublicpolicyortradeinterestsandissuesthroughstandardsprojects.Thesepartiesseeopportunitiesinthedraftingprocesstoencouragetheadoptionofpoliciesthatreflecttheiragendas.

2.Divergences:

?Regulatoryvs.voluntarystandards:HarmonisedEuropeanAIstandards,meaningthose

standardsthatareofficiallyalignedwiththeEUAIAct,provideaclearpathforthepresumptionofconformity.UnlessISO,IEEEandotherinternationaleffortsarerecognisedthroughthe

Europeanstandardisationbodies,theyremainvoluntaryintheEuropeancontext.Some

standardsfocusonlyonalimitedsetofusecases,particularlywhentheyareprimarilydesignedtodemonstratecomplianceofregulationfortheseusecases(e.g.high-riskEUAIActusecases).

?Relatedstandards:StandardsondatasharingordataintegrityneedtobecarefullycalibratedsoasnottoinadvertentlyacceleratefragmentationofAIstandards.

?Terminologyandscopedifferences:Variousbodiesusedifferentterminologiesand

methodologiestodefineAIrisks,transparencyandrobustnessaswellassecurityandsafety,

anditisunclearifsuchinconsistenciesmightleadtosignificantdifferencesinimplementationsacrosssectorsandjurisdictions.

Furthermore,standards-settingeffortsareoftenslowcomparedtothefast-pacedevolutionofAItechnologies,whichcouldleadtoagapbetweenemergingAIapplicationsandtheregulatoryortechnicalguidanceneededtoensuretheirresponsibleuse.

However,itisimportanttonote:

?Groundbreakingoranticipatorystandards(inanyfield)oftenrequiremoretimetodevelop,butcanbemoreresponsivethanregulation.

8

InternationalOrganisationforStandardisation(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IEC

42001:2023Informationtechnology—Artificialintelligence—Managementsystem(2023).

9

InternationalOrganisationforStandardisation(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IEC

23894:2023Informationtechnology—Artificialintelligence—Guidanceonriskmanagement(2023).

10

InternationalOrganisationforStandardisation(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IEC

42005:2025Informationtechnology—Artificialintelligence(AI)—AIsystemimpactassessment(2025).

July2025|AIgovernanceandstandards|7

?Standardsorganisationsoftenstartnewprojectstoaddressnewtrends,eventhoughexistingstandardsorminorupdatestothemcouldaddresstheseissues.

?Effortstocreatemutualrecognitionmechanisms,streamlinestandardswherepossibleand

ensurealignmentwithbroaderdigitalgovernanceframeworkswillbekeytofosteringamorecoherentandeffectiveAIstandardsecosystem.

Casestudies:CoreinternationalAIstandards

Market-drivenstandardsplayacrucialroleinAIgovernancebyestablishingcommonexpectationsacrosstheglobalAIsupplychain.Thecornerstoneofthesestandardsis

ISO/IEC42001,requiringorganisationsadoptingthisstandardforresponsibleAIto

implementmanagementpracticessuchasdemonstratingtheirabilitytoevaluateand

mitigaterisks,maintaininghigh-qualitydatadocumentationpracticesandensuringclear

communicationwithpartnersandcustomers.Additionally,organisationsusingor

developinghigh-riskAIsystemsmustimplementcontrolsatasystemleveldemonstrating

responsibleAIdesign,developmentanduse,suchas:Completionofsystemimpact

assessments;responsiblesystemlifecycledesignanddevelopmentanddataforAIsystems.Compliancecanbeverifiedthroughindependentaudits.

SupportingstandardscomplementISO/IEC42001,includingframeworksforassessingand

managingrisks(ISO/IEC23894),evaluatingpotentialimpactsontheorganisationand

individuals(ISO/IEC42005)andensuringdataqualitythroughouttheAIdevelopment

process(ISO/IEC5259-2)11.Thesestandardsservemultipleaudiences—fromtechnology

companiesseekingtobuildtrustworthyAIsystemstogovernmentagenciesdeveloping

policiesandorganisationslookingtoprocureAIsolutions.AsAIsystemsarefundamentallyITsystems,theymustbesecuredwithestablishedinformationsecuritypractices,suchas

ISO/IEC27001and27002,andISO/IEC27002,aswellasprivacypractices,suchasISO/IEC27701.

However,AI-specificthreatsalsoneedtobeconsidered.Aforthcomingstandard(ISO/IEC27090)willprovideAI-specificsecurityguidance.Inglobaltradeandsupplychaincontexts,especiallyforgenerativeAIandAIagents,structured,semanticdataplaysacriticalroleinensuringaccurate,safeandefficientoperations.TheICCDigitalStandardsInitiative(ICCDSI)hasadvancedthisthroughitsKeyTradeDocumentsandDataElements(KTDDE)12

modellingwork,basedontheUNCentreforTradeFacilitationandElectronicBusiness

(UN/CEFACT)’sUnitedNationsTradeDataElementsDirectory(UNTDED)ISO737213,which

standardisestradedocumentationandenhancesmachineinterpretability.UN/CEFACT’s

recentwhitepaperfurtherexploreshowAIcansupporttradefacilitationbyleveragingsuchstructureddataapproaches.14

IndustrycandemonstrateitscommitmenttoresponsibleAIbyadoptinginternational

standards,whilegovernmententitiesandpurchaserscanreferencetheminlegislation,

11

InternationalOrganisationforStandardisation(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IEC5259-

2:2024ArtificialIntelligence—DataQualityforAnalyticsandMachineLearning(ML)—Part2:DataQualityMeasures

(2024).

12

InternationalChamberofCommerce(ICC),DigitalStandardsInitiative,KeyTradeDocumentsandDataElements

(KTDDE)(2024).

13

InternationalOrganisationforStandardisation(ISO),ISO7372:2005TradeDataInterchange—TradeDataElements

Directory(June2005;reviewed2019).

14

UN,WhitePaperontheUseofArtificialIntelligencetoFacilitateTradeProcedures(2024).

July2025|AIgovernanceandstandards|8

policydevelopmentandprocurementprocessestomeetestablishedtrustandsecurityrequirementsbenchmarks.

AIriskmanagementformsacornerstoneofAIgovernance.Itreferstoframeworksthat

definepolicies,procedures,rolesandresponsibilitiesacrosstheAIlifecyclethat

organisationscanadoptinordertodevelop,deployandmaintainAIsystemsinawaythatminimisesrisksandattainsongoingregulatorycompliance.Implementationofsuchrisk

managementpracticeshasbeenmandatedunderseveralAIregulations.Leveraging

existingbestpracticereferencepointscanhelpdriveinteroperabilityamongdomesticAI

policyandregulationandacceleratetheimplementationofriskmanagementframeworks.ISO/IEC23894:2023AIRiskManagement,publishedinDecember2023,providesguidanceonhoworganisationscanmanagerisksspecificallyrelatedtoAIandisapplicablefor

organisationsofanysizeandacrosssectors.Inaddition,NIST’sAIRiskManagement

Framework(RMF)15,version1.0ofwhichwaspublishedinJanuary2023,supportsresponsibledevelopment,useandevaluationofAIproductsandservicesandispubliclyavailableatnocost.NISThasalsopublishedvariouscrosswalkstotheRMF,includingonetoISO/IEC42001:NISTCrosswalks16.

Althoughapproachestodetailedrequirementssuchasriskassessmentandmanagement

mayvaryacrossorganisations,adoptingvoluntaryconsensus-basedstandards(for

example,theextensiveworkofISO/IECJTC1SC4217,includingISO/IEC42001,ISO/IEC23894,ISO/IEC42005,ISO/IEC3850718)canserveasasolidfoundationformanagingAIrisks

throughouttheAIsystem’slifecycleandensureaninternationallyconsistentapproachtoimplementationofAIlaws.

3.Recommendations

ToensurethatAIstandardseffectivelysupportresponsibleAIgovernanceglobally,policymakersanddifferentstakeholdersshouldconsiderthefollowingrecommendations:

?PromotestrategicalignmentinAIstandards-development:EnsurethatAIstandardsare

developedinrelationtoidentifiedmarketneeds,commandstrongbusinesssupportanddonotconflictoroverlapwithwidelyusedstandards.

?Ensuredomestic/localbusinesses’andexperts’voicesarepartofAIstandardsdevelopment:

Giventhemanybenefitsofinternationalstandards,governmentsshouldraiseawarenessoftheopportunitytoinfluencemarket-drivenstandardsandencouragelocalexpertsfromalldomesticsectorstoparticipateinstandardsdevelopment,includingbusinessesthatdesign,developanddeployAIsystems.Industryexpertiseiscrucialforcreatingpractical,implementablestandards

thatalignwithtechnologicaladvancements,andlocalexpertiseiscrucialforshapingstandardswithlocalmarketrealities.

?Prioritiseindustry-drivenandgloballyrecognisedstandardsoverstrictlynationalorregional

regulatorycomplianceapproaches:Anindustry-led,internationalstandardfosters

interoperability,acceleratesinnovationandensuresthatstandardsremainpractical,adaptableandrootedinreal-worldapplications.

15

NationalInstituteofStandardsandTechnology(NIST),NISTAIRiskManagementFramework(AIRMF)(2023).

16

NationalInstituteofStandardsandTechnology(NIST),AIRMFCrosswalkDocuments(2023-2025).

17

InternationalOrganizationforStandardization(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IECJTC1/SC

42:ArtificialIntelligence—SubcommitteeonStandardizationintheAreaofArtificialIntelligence.

18

InternationalOrganizationforStandardization(ISO)/InternationalElectrotechnicalCommission(IEC),ISO/IEC

38507:2022Informationtechnology—GovernanceofIT—Governanceimplicationsoftheuseofartificialintelligenceby

organizations(2022).

July2025|AIgovernanceandstandards|9

?ChampionmultistakeholdercollaborationinAIstandardisation:Governmentsshouldpromoteengaginginstandardsandadvocatefortheirbenefitswithintheircountriesandinmulti-and

bi-lateraltalks.AIstandardsshouldbedevelopedthroughtransparent,inclusiveand

multistakeholderprocessesthatinvolvevolunteersfromallsectors:industryleaders,academia,civilsocietyandpolicymakers.Thisensuresthatstandardsarerobust,balancedandreflectiveofdiverseperspectives,enhancingtheirlegitimacyandadoption.AIisaglobaltechnology,

andregulatoryfragmentationcanhinderinnovationandcross-bordercollaboration.Atthesametime,fragm