TheFutureof

AIGovernanceTheUAECharterandGlobalPerspectivesREPORTincollaborationwith2Principle

1:

Strengthening

Human-Machine

Ties

10Principle

2:

Safety

14Principle

3:

Algorithmic

Bias

18Principle

4:

Data

Privacy

22Principle

5:

Transparency

26Principle

6:

Human

Oversight

30Principle

7:

Governance

and

Accountability

34Principle

8:

Technological

Excellence

38Principle

9:

Human

Commitment

42Principle

10:

Peaceful

Coexistence

with

AI

46Principle

11:

Promoting

AI

Awareness

for

an

Inclusive

Future

50Principle

12:

Commitment

to

Treaties

and

Applicable

Laws

54Table

of

Content3Table

of

ContentsTopicsThe

UAE

Charter:

The

12

AI

Principles

6KPMG’s

Trusted

AI

Framework

8147Werecognizethataclear,actionablesetof

AIprinciplesformsthecornerstoneof

ethicalandresponsibleAIdevelopment.Theseprinciplesarenotonlyessentialforbuildingpublictrustandensuringorganizationalaccountability,butalsoforfosteringinclusiveinnovationthat

benefitscitizens,businesses,andgovernmentsalike.Asglobalregulatoryframeworksevolved,suchastheEUAIAct

passedin2024,groundedintheEuropeanCommission’sethical

guidelinesfortrustworthyAI,principles-basedgovernancehasemergedasthefoundationalapproachtoAIoversight.TheUAEhasdemonstratedregionalandgloballeadershipthroughitsAIStrategy2031andthereleaseof

theUAEAICharterforthedevelopmentanduseof

ArtificialIntelligence,

inJuly2024,whicharticulatestwelvekeyprinciplestoensure

AIisdeployedsafely,equitably,andtransparently.Thiswhitepaperoffersadetailedinterpretationof

eachof

the

12UAEAICharterprinciples,actionablerecommendationsforimplementationacrosstheAIlifecycle,mappedtoKPMG’sTrustedAIFramework,practicalinsightstosupport

AIgovernance,riskmanagement,andregulatoryalignment,andablueprintforbuildingresilient,human-centricAIsystemsinalignmentwiththeUAE’snationalpriorities.TheUAECharterplacesparticularemphasison

humanoversight,inclusivity,safety,andlegalcompliance—valuesthatresonatewithglobalAIethicsstandardslikethoseoutlinedbyOECD,UNESCO,andtheEU.AsAI

regulation

becomes

more

stringent,organizationsthatproactivelyalignwiththeseprincipleswillbebetterpositionedtoleadresponsibly,mitigate

risks,andcapturethefullpotentialof

AIinnovation.Tomovebeyondaspirationalintent,organizationsmustembedtheseprinciplesintooperationalreality.Thismeansevolvingexistinggovernancemodelstosupportthedistinctrequirementsof

AI—suchasdataprovenancetracking,model

accountability,explainability,biasaudits,andhumanoversight.Governanceframeworksmustshiftfromstaticpoliciestoadaptivecontrolsthatalignwiththefast-evolvingAIlifecycle.ForewordEmbeddingtheUAEAICharterintoenterprisegovernancealsoprovidesastrategicadvantage.It

signalsreadinessforfuturecompliance,enablesrisk-awareinnovation,andensuresthatAIdeploymentsarenotonlylawful

butalso

aligned

withpublicexpectationsandsocietalvalues.Organizationsthatoperationalizetheseprinciplesearlywillbebetterequippedto

manageethicaldilemmas,respondtoregulatoryinquiries,andbuildlastingtrustwithusers,regulators,andthewider

community.Proactivelyimplementingtheseprinciplesnotonlyensuresregulatoryreadiness

but

alsodeliversclearbusinessvalue.OrganizationsthatembedresponsibleAIpractices

early

canaccelerateinnovationwithconfidence,reducecompliancecosts,andenhancetheirreputationas

trustworthy,forward-thinking

leaders.BybuildingAIsystemsthataretransparent,inclusive,andhuman-centric,businessescanunlocknewopportunities,gainstakeholdertrust,anddifferentiatethemselvesinanincreasinglyAI-driven

economy.Acrosstheglobe,

jurisdictionssuchastheEuropeanUnion,Canada,theUnited

States,andSingaporearemovingswiftlytocodifyAIethicsintobindinglegislationandoperationalframeworks.Thissignalsaglobalshiftwhere

AIgovernancewillnolonger

beoptional—butacorecomponentofdigitalcompetitivenessandenterpriseresilience.51.StrengtheningHuman-MachineTies:TheUAEaimstoenhancetheharmoniousandbeneficial

relationship

betweenAIandhumans,ensuringthatallAIdevelopmentsprioritize

humanwell-beingandprogress.2.Safety:The

UAEplacesgreatimportanceonsafety,ensuringthatallAI

systems

complywiththehighestsafetystandards.

Thecountryencouragesmodifyingorremovingsystemsthatpose

risks.△3.Algorithmic

Bias:TheUAEaimstoaddressthechallengesposedbyAIalgorithms

regarding

algorithmicbias,contributingtoafairandequitableenvironmentforallcommunitymembers.

ThispromotesresponsibledevelopmentofAItechnologies,makingtheminclusiveandaccessibletoeveryone,supportingdiversity,andrespectingindividualdifferences.Itensuresequaltechnologicalbenefitsandimprovesqualityoflifewithoutexclusion

ordiscrimination.4.DataPrivacy:In

linewiththe

UAE’sstanceon

privacy

rights,whiledataisessentialfor

AIdevelopment,supportingandpromotinginnovationinAI,the

privacy

ofcommunitymembersremainsatoppriority.5.Transparency:The

UAEseekstocreateaclearunderstandingofAIandhow

systems

operate

andmakedecisions,whichhelpsbuildtrust,enhance

responsibility,andpromoteaccountabilityintheuseof

thesetechnologies.6.

HumanOversight:TheCharteremphasizestheirreplaceablevalueofhuman

judgmentand

humanoversightoverAI,aligningwithethicalvaluesandsocialstandards

tocorrectanyerrorsor

biasesthat

mayarise.TheUAECharter:The12AIPrinciples6

WorldGovernmentsSummit7.

GovernanceandAccountability:TheUAEadoptsaresponsibleandproactivestance,

emphasizingtheimportanceofgovernanceandaccountabilityinAItoensurethetechnology

isusedethicallyandtransparently.8.

TechnologicalExcellence:AIshouldbeabeaconofinnovation,reflecting

the

UAE’svision

ofdigital,

technological,andscientificexcellence.

TheUAEseeksgloballeadership

byadoptingtechnologicalexcellenceinAItodriveinnovation,enhance

competitiveness,andimprovequalityoflifethroughinnovativeandeffectivesolutionstocomplexchallenges,contributingtosustainable

progressbenefitingsocietyasawhole.9.

HumanCommitment:HumancommitmentinAIreflectsthespiritofthe

UAE,essentialforensuringthatthedevelopmentof

thistechnologyservesthepublicgood.

Itfocusesonenhancinghumanwell-beingandprotectingfundamentalrights,emphasizingtheimportanceofplacinghumanvaluesattheheartof

technologicalinnovationtoensureapositiveandlasting

impacton

society.10.

PeacefulCoexistencewithAI:PeacefulcoexistencewithAIiscrucialtoensuretechnologyenhancesthe

well-beingandprogressofourcommunitieswithoutcompromisinghuman

securityorfundamentalrights.11.

PromotingAIAwarenessforanInclusiveFuture:It

is

essential

to

create

an

inclusive

future

that

ensures

everyone

canbenefitfromAIadvancements,guaranteeingequitableaccesstothis

technologyanditsadvantagesforallsegmentsofsociety.12.

CommitmenttoTreatiesandApplicableLaws:TheUAEemphasizestheimportanceofcomplyingwithinternational

treatiesandlocallawsinthedevelopmentand

use

ofAI.7Asartificialintelligencebecomesincreasinglyintegraltocriticaldecisionsandeverydayoperations,

KPMGdevelopeditsTrustedAIFrameworktohelporganizationsnavigatethisevolvinglandscape.Theframeworkbringsstructure,accountability,andclaritytotheAIlifecycle,ensuringthatAIsystemsareethical,transparent,andalignedwithhumanvaluesfromstrategytodeployment.BuiltonKPMG’sglobalexperienceacrossindustries,theframeworkisfoundedontencoreprinciples.Theseprinciplesincludefairnessandtransparency,whichensureAIsystemsareinclusiveandunderstandable;explainabilityandaccountability,whichfosterhumanoversightandresponsibility;andprivacy,security,and

safety,whichprotectbothindividualsandsystems.Additionally,theframeworkemphasizesdataintegrityandreliability

forconsistentAIperformance,aswellassustainabilitytoensureAIadvancementscontributetobroader

socialandenvironmentalgoals.KPMG’sTrusted

AIFramework8Accountability,Transparency,DataIntegrity12.Commitment

to

Treaties

and

Applicable

Laws

Accountability,

Privacy,

Data

IntegrityThisclosealignment

betweentheUAEAICharterand

KPMG’sTrusted

AI

Framework

provides

astrongfoundationforaction.TheTrustedAIprincipleshavealready

beenoperationalized

throughdefinedmethodologiesacrosstheAIlifecycle—spanningstrategyanddesign,

dataenablement,modeldevelopment,testingandevaluation,anddeploymentandmonitoring.Buildingonthisprovenfoundation,thesamestructuredapproachhasbeenappliedinthis

whitepapertotheUAE’stwelveAIprinciples.Foreach,

practicalsteps

are

outlinedto

helporganizationsembedethical,human-centricAIpracticesandturnprinciplesintotangibleoutcomes.98.Technological

ExcellenceReliability,Sustainability9.HumanCommitmentFairness,Sustainability,Accountability10.Peaceful

Coexistence

with

AISafety,Security,

Fairness11.PromotingAIAwarenessforanInclusiveFutureFairness,ExplainabilityTheUAEAICharterreflectsasimilarcommitmentto

responsibleAI

development,

expressing

a

nationalvisionthroughtwelveguidingprinciplesthataligncloselywiththoseinKPMG’sTrusted

AI

Framework.ThetablebelowillustrateshoweachUAEAI

principle

mapsto

one

or

more

ofKPMG’sTrustedAI

principles:UAEAIPrincipleAlignedKPMGGlobalTrustedAIPrinciple(s)1.Strengthening

Human-Machine

Ties

Explainability,Fairness,Accountability2.SafetySafety,

Reliability,Security3.Algorithmic

BiasFairness,Transparency,

DataIntegrityGovernance

and

Accountability6.HumanOversightAccountability,ExplainabilityTransparency,

ExplainabilityPrivacy,Data

Integrity4.

Data

Privacy5.Transparency7.StrengtheningHuman-MachineTiesTheUAEaimstoenhancethe

harmoniousandbeneficialrelationship

between

AI

andhumans,ensuringthatallAIdevelopmentsprioritizehuman

well-beingandprogress.Principle

110UnderstandingthePrinciple

inReal-WorldTermsThisprincipleaimstoensurethatAIsystemsenhance

andaugmenthumancapabilities,empoweringhumanbeingstoexceedtheirpotential

bycreatingsmarter,moreinclusivesolutions.AIshouldalignwithethicalprinciples,respectinghuman

dignity,rights,andvalues.Ultimately,theUAEaimstofoster

anenvironmentwherehumansandAIcollaboratetoimprovequalityoflife,boostproductivity,anddrive

societalprogress.Real-worldexamples:?

HealthcareAI:AIsystemsusedin

healthcare

toassistdoctorsindiagnosingdiseases

moreaccuratelyandefficiently,ultimatelyimprovingpatientoutcomes.?SmartCities:

AI-driventechnologiesintegratedintourbanplanningtoimproveinfrastructure,optimizetrafficflow,andenhancethequalityoflifeforresidents.EmbeddingThisPrincipleintoAIGovernanceTostrengthenhuman-machineties,focusondevelopingAIsystemsthataugmenthumancapabilities,enhancewell-being,anddrivepositiveoutcomesforemployees,customers,andsociety.EnsurethatyourAIinitiativesalignwithbestpracticesandthoughtfullyconsidertheirbroaderimpactonhumanvalues,dignity,andrights,whilealsoreflectingtheculturalandsocietalvaluesof

theUAEthroughouteverystageofdevelopmentandimplementation.Consider

incorporating

human-in-the-loopdecision-making

to

further

reinforce

the

human-AIrelationship,ensuringmeaningfuloversight,trust,andaccountability.11?HumanImpactAssessment:Assessthepotentialpositiveandnegativeimpactsof

AIsystemsdevelopedonhumanwell-being,ensuringtheoutcomesarealignedwiththeintendedbenefits.?

UserFeedback:Incorporatefeedbackfromuserstofine-tuneAIsystems,ensuringtheyarerelevanttohumanneeds

and

progress.DeploymentandMonitoring

?

ContinuousCollaboration:

MaintainactivecollaborationwithhumanusersandstakeholderstoensurethatdeployedAIsystems

remainbeneficialandenhancehuman

progress.?

MonitorAIforHumanImpact:Trackthelong-termeffectsof

AIonsociety,ensuring

AIsystemscontinuetoprioritizeandenhance

human

well-being.?

AdaptationtoHumanNeeds:Continuouslyadapt

AItechnologiestomeettheevolvingneedsand

valuesofhumanusers,especiallyassocietalcontextschange.?

DataSensitivity:

Buildtrustbyensuringdatacollectionandprocessingrespects

humanprivacyanddignity,ensuringtheresponsibleuse

ofpersonalandsensitivedata.?

Well-beingMetrics:Considerfactorssuchaswell-being,safety,anduserexperiencewhenevaluatingthedatausedfortrainingAImodels.?DiverseDataRepresentation:

Usedatasetsthat

reflectdiversehumanexperiences,ensuringAIsystemscanservethebroadspectrum

of

societal

needs.ModelDevelopment

?Human-AICollaborationFeatures:

DevelopAIsystemsthatenhancehumancapabilities

byoffering

insights,and

providing

support,without

replacinghumandecision-making.?

Human-CentricDesign:

DesignAIsystemstoaugmenthumancapabilitiesbycontinuouslygatheringdiversefeedbackandusingittorefineandenhanceAI’simpact.?

EthicalAIGoals:SetclearethicalguidelinesforAI

development

that

prioritize

human

well-beingandaddresspotentialnegativeimpacts,

ensuringalignmentwithbothglobalstandards

andUAE’s

culturalvalues.?

Human-in-the-LoopIntegration:Considerembedding

human-in-the-loop

mechanisms

earlytostrengthendecision-making,ensure

accountabilityandalignAIsystemswithcore

humanvalues.?

TransparentAlgorithms:

Build

modelsthatallowhumanstoeasilyunderstand,trust,andcollaboratewithAIsystems.Transparencyhelps

ensurehumanoversightismaintained.?BiasReduction:

EnsureAI

modelsarefreefrombiasesthatmay

harm

human

progress,includingensuringequitabletreatmentacross

diversegroupsandpreservingsocietalvalues.Principle

1BestPracticesandMethodologiesTestingandEvaluationStrategyandDesignDataEnablement12KeyTools,TechniquesandFurtherReadingToolsandTechniques:?

Human-CenteredAI

Design

Frameworks?

Human-AICollaborationToolkits

(e.g.Microsoft

Copilot,Salesforce

Einstein)?UX

ResearchandCognitive

LoadTestingtools

(e.g.OptimalWorkshop)?KPMGTrustedAI

Framework?KPMGTrustedAI

RiskandControl

Matrix

(RCM)ExtendingthePrincipleto

AgenticAISystemsAgenticAIsystemsmust

bedesignedtocomplement,notreplace,

human

roles.Theyshouldenhancehumandecision-makingandproductivitythroughcontextualawarenessandfeedbackmechanisms.Ensuringintuitivehumaninteractionandtransparencywillhelppreservetrust.Organizationsmustprioritizeuser

experienceinagent-AIinterfaces.Emotionalandcognitiveimpactonusersshouldbe

monitoredandimproved

overtime.FurtherReading:?

StanfordHAI:Human-AICollaborationStudies?Harvard

Berkman

KleinCenter:

Ethicsof

Augmentation?Microsoft:The

FutureComputed–AIand

Human

ValuesPrinciple2SafetyTheUAEplacesgreatimportanceonsafety,ensuring

thatall

AIsystemscomply

with

thehighestsafetystandards.

Thecountryencouragesmodifyingorremovingsystems

thatposerisks.14UnderstandingthePrinciple

inReal-WorldTermsAIsafetyreferstoensuringthatAIsystemsfunctionasintended,withoutcausingharmtoindividuals,businesses,orsociety.Thisincludestechnicalrobustness,risk

mitigation,and

incorporating

fail-safestopreventoraddressunintendedconsequences

orfailures.The

EUAIActexemplifiesthisapproachbymandatingstringentsafetystandardsforhigh-riskAIapplications.Prioritizingsafetyisessentialtominimizingrisksandmaintainingboth

operationalcontinuityandpublictrustintheUAE.Real-worldexamples:?

AutonomousVehicles:AI-drivencarsfailingtorecognizepedestriansinlowvisibilityconditions,leadingtoaccidentsandregulatoryscrutiny.?

HealthcareAI:

DiagnosticAImisinterpretingmedicalimages,leadingtoincorrecttreatments

and

potential

liability

risks.EmbeddingThisPrincipleintoAIGovernanceEnsuringAIsafetyrequiresastructuredapproach—

fromriskassessmentstocontinuoustestingandfail-safemechanisms.Toolslike

KPMG’sTrustedAI

Risk

Frameworksupportthisprocessbyoffering

astructuredmethodologytoidentify,assess,andmitigate

AI-related

risks,including

those

tied

tosafety,inalignmentwithstandardssuchasISO42001andthe

EUAIAct.CombinedwithastrongAIgovernanceframework,thesetoolshelpensuresafetymeasuresremaineffective,transparent,andalignedwithbothlocalandglobal

best

practices.

By

embeddingsafetybestpracticesintoeverystage

of

development,organizationscanenhancereliability,maintain

compliance,and

build

trust.15?Adversarial

Testing:

Identify

weak

points

intheAIsystembytestingagainstpotentialfailurepointsandestablishcorrectivemeasures

before

deployment.?

TestingforEdgeCases:

EvaluateAI

performanceunderextremeconditions(e.g.lowvisibilityforself-drivingcarsorunpredictablemarketfluctuations

in

finance).?

SafetyBenchmarking:

Defineand

measuresafetyperformanceagainstindustrystandards.DeploymentandMonitoring

?

ContinuousSafetyMonitoring:

Regularlyaudit

AI

systems

post-deployment

to

detectanomaliesorfailures.?

IncidentResponsePlans:

EstablishclearescalationprotocolsforAImalfunctionsto

ensurequickremediation.?RegulatoryComplianceReporting:

Documentandcommunicatesafetymeasureswithintheorganizationtodemonstrateadherencetosafety

standards.?

DataIntegrityChecks:Validatetrainingdata

foraccuracy,completeness,andconsistencytopreventAIfailures.?BiasandAnomalyDetection:Identify

biases

thatcouldleadtounsafeAI

behavior,

such

asmisclassificationinhealthcareorautonomoussystems.?

SimulationandStressTestingData:Train

AI

modelsonvariedscenarios,includingedge

cases,toensurerobustnessinreal-worldapplications.ModelDevelopment

?

Safety-ConsciousAlgorithms:Implementalgorithmsthatprioritizesafety,incorporating?

SafetyGoalsandMetrics:

Establishclear

safetygoalsandmetricsforAIinitiatives,focusingonreliability,resilience,transparency,andsecurity.

Tools

like

KPMG’sAI

metricscan

measureperformanceandensureethicalalignment.?

RiskIdentification:

Definesafety

risksassociatedwiththeAIsystemsandestablishprotocols

for

risk

mitigation.?

StakeholderConsultation:

Engage

regulators,industryexperts,andend-userstoanticipate

safetyconcernsbeforedevelopment.?

Fail-SafeDesign:

EnsurethatthedesignofAIsystemshasclearoverride

mechanismstopreventharmincaseof

failure.Incorporatefallbackmechanisms,monitoring,andhuman-

in-the-loop.guardrailsandconstraintstopreventharmfuldecisions.?

Fail-SafeMechanisms:

Embedfail-safemechanisms

like

human-in-the-loop

andlogginginthefinaldesign.?ExplainabilityandTransparency:

EnsurethatAIdecisionscanbe

understoodandauditedtoidentifypotentialsafetyrisks

beforedeployment.Principle2BestPracticesandMethodologiesTestingandEvaluationStrategyandDesignDataEnablementKeyTools,TechniquesandFurtherReadingToolsandTechniques:?

AdversarialTesting

Frameworks

(e.g.

CleverHans,

Foolbox)?FormalVerificationTools

(e.g.TLA+,Z3)?Bayesian

Networks,

MonteCarlo

Dropout?RedTeamingandSimulation

Labs?KPMGTrustedAI

Framework?KPMGTrustedAI

RiskandControl

Matrix

(RCM)ExtendingthePrincipleto

AgenticAISystemsAgenticAIintroducesdynamicdecision-making,whichrequiresreal-timerisk

detection

andmitigationcapabilities.Safetyprotocolsmust

beembeddednot

justincode,butalsoin

how

agentsinteractwithsystemsandpeople.Fail-safesandescalationpathstohumansupervisorsareessential.Simulationtestingforadversarialorunintendedagentbehaviormustbe

prioritized.

Organizations

shouldtrackagentactionstoensureaccountability.FurtherReading:?NISTAI

Risk

Management

Framework?

OpenAI’sSystemSafety

Practices?EUAIAct:Safety

Provisionsfor

High-Risk

SystemsPrinciple3AlgorithmicBiasTheUAEaimstoaddressthe

challenges

posed

byAIalgorithmsregardingalgorithmicbias,contributing

to

a

fair

and

equitableenvironment

forallcommunity

members.

ThispromotesresponsibledevelopmentofAI

technologies,making

them

inclusive

and

accessible

toeveryone,supporting

diversity,

andrespectingindividualdifferences.It

ensures

equaltechnologicalbenefitsandimprovesqualityoflife

without

exclusionordiscrimination.18UnderstandingthePrinciple

inReal-WorldTermsInpractice,algorithmicbiasoccurswhen

AIsystemsmakedecisionsthatunintentionallyfavorordisadvantagecertaingroupsbasedonfactors

like

gender,ethnicity,age,orsocioeconomicstatus.This

canstemfrombiasedtrainingdata,flawedmodelassumptions,oralackofdiverserepresentationin

development.Addressingbiasiscriticaltobuildingtrust,ensuringfairness,andmitigatingfinancial,legal,andreputational

risks.Real-worldexamples:?

HiringSystems:AIsystemsrejectingfemalecandidatesbasedonbiasedtrainingdata

derivedfrom

male-dominated

industries,exposing

thecompanytodiscriminationclaimsorregulatorypenalties.?

LoanApprovals:AI-basedcreditsystemsrejecting

loanapplicantsbasedonhistoricaldiscriminatorypractices,potentiallyviolatingfairlendinglaws.EmbeddingThisPrincipleintoAIGovernanceToeffectivelyaddressalgorithmicbiasinyourAIsystems,embedfairnessintoeveryphaseofdevelopment.Thisincludesmakingproactivedesign

choices,usingrepresentativeandbalanced

data,

andconductingcontinuoustestingtoensureequitableoutcomes.

EffectiveAIgovernance,supportedbyastrongframework,shouldbewovenintoeachstageto

ensureaccountability,transparency,andcompliancewithethicalandlocalregulatorystandards.

By

doingso,organizationscantranslatetheprincipleoffairnessintoactionable,impactfulstepsthatdriveresponsibleAIdevelopmentthatalignswiththeUAE’s

requirements.19?

ThresholdSetting:

Beforedeployment,defineacceptablefairnessthresholdsthatalignwiththefairnessgoalsandmetricssetattheideation

stage.?ImpactTesting:

Testthefullytrained

modelsfor

biasagainstthefairnessthresholdsandevaluate

howtheAIsystem’soutcomesdifferacrossdemographicsandadjustasnecessarytoensureequitableresults.DeploymentandMonitoring

?

OngoingMonitoring:Continuouslymonitortheperformanceof

AIsystemspost-deployment