ADCM

Academy

Researchcentre

AIApplicationsinWeb3SupTechandRegTech:

ARegulatoryPerspective

Preparedby

ShengliangLu*,AmritpalSidhu*,BingshengHe^,BrianByagaba*,

DmitryFedotov*,AlejandroVargas#,HelenaZhang,BarryWest*

FULLPAPER

ADGM(includingFSRARA

andAcademy)NUS,#Mooncheck

Abstract

ThedigitalrealmisexperiencingatransformativeshiftdrivenbytheemergenceofWeb3technologies

andvirtualassets.Thisnewphaseofinternettechnologyleveragesdistributedledgertechnologiesand

smartcontractswhilefosteringdecentralization,increasingtransparency,andreducingdependence

onintermediaries.SuchinnovationsarepivotalinshapingDecentralizedFinance.However,therapid

adoptionofWeb3technologiespresentssignificantrisksunderscoredbyhigh-profilefailuresandsystemicvulnerabilities.

ADGMhasdevelopedaconduciveregulatoryenvironmentthroughitsFinancialServicesRegulatoryAuthority(FSRA)bycreatingatransparentandprogressiveregulatoryframeworkalignedwith

internationalstandardsandsafeguardingstakeholders?interests.

ThiswhitepaperexplorestheintegrationofAIintoregulatorytechnologiestoenhancecompliance

monitoringandriskmanagement.ItdetailstheresearchanddevelopmenteffortsbytheNational

UniversityofSingapore?sAsianInstituteofDigitalFinance,ADGMFSRA,andtheADGMAcademyResearchCentre.Thewhitepaperconcludeswithasummaryofkeyfindingsandproposesfuturecollaborative

effortstofurtherdeveloptheregulatorylandscape.

Keywords:

AI,Cryptocurrency,DeFi,DigitalAsset,RegTech,Regulation,SupTech,VirtualAsset,Web3

ADGM

2

TableofContents

Abstract 2

1–Introduction 4

2–Background 6

3–OpportunitiesforUtilisingAIforRegulatingWeb3Activities 7

4–AIInnovations 8

4.1EmergingAITechnologies

9

4.2AISolutionsintheWeb3RegulatoryDomain 10

4.3ChallengesinAIImplementation 11

4.4FutureDirections 12

5–PilotsofAIInnovationsinADGM(JointWorkwithNUSAIDF) 13

5.1Pilot1:SmartContractSuitabilityAssessmentwithAI 13

5.2Pilot2:AuditReportAssessment 15

5.3Pilot3:SmartDueDiligencewithAI 16

6–ConclusionandFutureWork 18

6.1Conclusion 18

6.2KeyTakeaways 18

6.3FutureWork 19

ACKNOWLEDGMENTS 20

DISCLAIMER 22

ADGM

3

1–INTRODUCTION

ThedigitallandscapeisundergoingrapidtransformationwithWeb3technologiesleadingadvancementsininternettechnology.Builtondistributedledgertechnologies(DLT)andsmartcontracts,Web3technologies

emphasizedecentralization,increasetransparencyandreducerelianceonintermediaries.DLT,including

blockchains,offersasecure,immutableledgerfortransactionsanddata,whilesmartcontractsfacilitate

automatedagreementswithoutintermediaries.Thiscombinationsupportsthedevelopmentofdecentralizedapplications,particularlyinDecentralizedFinance(DeFi),whicharereshapingfinancialtransactionsthroughpeer-to-peerinteractions.

Theglobalcryptocurrencymarketcapitalisationhassurpassedthe$3trillion1markrivallingsomeofthe

world’slargestcompanies,includingAppleandMicrosoft2.Thecryptocurrencyuserbasehasexpanded

significantly,witha34%increasein2023alonerisingfrom432millioninJanuaryto580millionbyDecember3.Thisgrowthunderscorestheincreasingadoptionandintegrationofcryptocurrenciesintotheglobalfinanciallandscape.Additionally,datarevealsthattheUnitedArabEmirates(UAE)leadstheworldincryptocurrencyadoption,withover30%ofitspopulation,approximately3millionpeople,owningdigitalassets4.Thisreflectsthenation’sforward-thinkingembraceoffinancialtechnologyanditsaspirationstobecomeamajorfintechhub.

the

but

ADGMplaysapivotalroleintherapidlyevolvingfinanciallandscape.Overseeingfinancialservicesintheinternationalfinancialcentreandfreezone,theADGMFinancialServicesRegulatoryAuthority(FSRA)isatforefrontoffosteringaregulatoryenvironmentthatisnotonlyseekstosupportthegrowthofDeFiandVAalsothebroaderdigitaltransformationinfinancialservices.

and

TheFSRAhasdevelopedacomprehensiveregulatoryframeworkforvirtualassetsthathasbeencontinuallyrefinedsinceitsintroductionin20185.Thisframeworksupportsinnovationwhileensuringrobustoversight

alignmentwithinternationalstandards.Byembracingdigitaltransformation,ADGMcollaboratescloselywithtechnologyecosystempartnerslikeHub716andresearchinstitutesliketheNationalUniversityofSingapore,promotingtheadoptionofcutting-edge,technicalsolutionswithinADGM.Thisproactiveapproachhelps

positionAbuDhabiasadestinationofchoiceforfinancialcompaniesseekingtoleverageadvancedtechnologiesanddigitalfinancemodels.

Furtherenhancingitsregulatorycapabilities,theADGMFSRAisleveragingadvancementsinRegTechandSupTechtostreamlineregulationandsupervision.ThroughAI-drivenRegTechsolutions,theFSRAcould

providemoreinteractiveandtailoredregulatoryinteractions,makingcompliancemoreefficientand

accessibleforentitiesoperatingwithinADGM.TheimplementationofAI-enabledSupTechtoolscould

supporttheFSRA’soversightandriskmanagementobjectiveswhilereducingcostsforfinancialinstitutions.

Together,theseinitiativesunderscoretheFSRA’smandatetoprovideatransparent,efficient,andprogressivefinancialenvironmentthatnotonlysafeguardstheinterestsofcustomers,investors,andindustryparticipants,butalsofosterssustainablegrowthandinnovationinADGM.

1AsofNovember2024,theglobalcryptocurrencymarketwasat$3.5trillion.

/digital-assets/crypto-prices

2

https://8/

3

/company-news/global-cryptocurrency-owners-grow-to-580-million-through-2023

4

/uae-dominates-global-crypto-adoption-vietnam-surges-to-second

5ADGM(2018)|ADGMlaunchescryptoassetregulatoryframework

/media/announcements/adgm-launches-crypto-asset-regulatory-framework

6Hub71|Hub71

ADMC

Academy

Researchcentre

ADGM

4

SupTech(SupervisoryTechnology)referstotheapplicationoftechnologytoenhancethesupervisory

andoversightfunctionsofregulatoryauthorities.Itinvolvestheuseofadvancedtoolslikedataanalytics,AI,andautomationtoimprovethemonitoringandsupervisionofregulatedactivities,andenforcement

ofregulatoryframeworks.SupTechaimstoprovideregulatorswithmoreeffective,data-driveninsights,andenablethemtobetteridentifyissues,assessrisks,andenforceregulationsinreal-time.

RegTech(RegulatoryTechnology)referstotheuseoftechnologytostreamline,automate,andimproveregulatorycomplianceprocessesforbusinesses.ItleveragesinnovativetoolssuchasAI,machine

learning,automation,anddataanalyticstohelpcompaniesmeetregulatoryrequirementsmore

efficiently,reducecompliancecosts,andenhancetransparencyandreporting.RegTechaimstosimplifycomplexcompliancetasks,suchasmonitoringtransactions,identifyingrisks,andensuringadherencetolegalstandards

NewrisksarisingfromthenatureofWeb3technologies,suchasfailuresofblockchainprotocolslikeTerra(LUNA)7,alongsideemergingvulnerabilitiesinsmartcontracts8,stresstheneedforeffectiveregulatory

frameworksandriskmanagementstrategies.Theinnovativeanddecentralizednatureofblockchain

technologycreatesafertilegroundofexposuretonewformsoffraudandsystemicfailuresthatmustbeaddressedforwideradoptiontobepossible.

Inresponse,amongotherstrategies,ADGMisexploringtheapplicationofAIinregulatoryandsupervisory

technologysolutionstoimprovecompliancemonitoringandriskmanagement.TheAsianInstituteofDigital

FinanceattheNationalUniversityofSingapore(NUSAIDF)conductsFinTechresearchinAItechnologies,

providingtoolsforpredictiveanalytics,anomalydetection,andautomatedcompliance.TheFSRAistesting

andvalidatingtheseAItechnologiestomeettheemergingneedsofregulatingandsupervisingWeb3andVAecosystemseffectively.ThiswhitepapersummarizestheresearchanddevelopmenteffortsofNUSAIDFand

ADGM(includingboththeFSRAandtheADGMAcademyResearchCentre)inimplementingAItechnologiestosupportregulatoryandsupervisoryactivitiesintheWeb3andvirtualassetdomains.

Asthispaperisintendedforabroaderaudienceandnotsetuptoprovidespecificdefinitionsreadersshouldnotethattheterms“VirtualAssets”,“Web3”,“blockchain”,“DLT”,and“network”areusedinterchangeably.

Nonetheless,explanationsofsometermsareprovidedinSection2.

Theremainderofthepaperisoutlinedasfollows.Section2providesthecontextandscopeofthispaper

whileSection3discussesthepotentialopportunitiesforregulatorstoleverageAItechnologies.Section4

discussesAIinnovationsshapingregulatoryactionsandactivities.Section5examinespilotprojects

conductedbyNUSAIDFandADGM,demonstratingpracticalapplicationsoftheseinnovations,suchassmartcontractassessments,securityauditing,andAI-poweredduediligence.Section6concludesthepaperwithasummaryoffindingsandadiscussiononfuturedirectionsandpotentialareastoenhancetheregulatory

landscape.

7Investopedia(2022)|TerraUSDCrashShowsRisksofAlgorithmicStablecoins

/terrausd-crash-shows-risks-of-algorithmic

-stablecoins-5272010

8Cointelegraph(2023)|Curve-Vyperexploit:Thewholestorysofar

/news/curve-vyper-exploit-whole-story-so-far

ADM

Academy

Researchcentre

ADGM

5

2–BACKGROUND

Thissectionaimstoprovideexplanationsforkeytermsthatareusedinthispaper,settingafoundationalbackgroundforreaderstobetterunderstandthediscussionsinlatersections.

VirtualAsset.TheFSRA’sregulatoryframeworkcategorizesdigitalassetsintodistinctclasses,whichalsoincludeFiat-ReferencedTokensandDigitalSecurities.

VAreferstoadigitalrepresentationofvaluethatcanbedigitallytradedandfunctionsas(1)amediumofexchange;and/or(2)aunitofaccount;and/or(3)astoreofvaluebutdoesnothavelegaltenderstatusinanyjurisdiction.AVAis(a)neitherissuednorguaranteedbyanyjurisdictionandfulfilstheabovefunctionsonlybyagreementwithinthecommunityofusersoftheVA;and(b)distinguishedfromfiatcurrencyand

e-money9

Web3.Web3representsthenextevolutionoftheinternet,transitioningfrom“read”(Web1)and“read-

write”(Web2)to“read-write-own”capabilities10.UnlikethecentralizedplatformsofWeb2,Web3leverages

blockchaintechnologytoempoweruserswithtrueownershipoftheirdata,digitalassets,andonline

interactions.Thisdecentralizedparadigmreducesrelianceonintermediaries,fosteringgreateruserautonomyandprivacywhileredefininghowindividualsinteractwithdigitalplatforms.

DLTandBlockchainNetwork.DLTreferstoadigitalsystemforrecordingtransactionsofassetsinwhichthe

dataisstoredacrossmultiplesitesornodessimultaneously.Unliketraditionalcentralizeddatabases,DLTsaredecentralized,eliminatingtheneedforacentralauthorityandenhancingtransparencyandsecurity.Each

participantinthenetworkmaintainsasynchronizedcopyoftheledgertherebyreducingtheriskofsinglepointsoffailure.

BlockchainisaspecifictypeofDLTthatorganizesdataintocryptographicblockswhicharethenlinkedchronologicallytoformachain.Thisstructureensuresthatrecordeddatabecomesimmutable.VAsaretypicallybuiltonblockchainnetworks.InWeb3,DLTandblockchainnetworkspowerDeFiplatformsanddecentralizedapps(dApps)byenablingsecure,transparenttransactions.

DeFi.DeFireferstoafinancialecosystembuiltonblockchainandDLTthatenablespeer-to-peer

transactionsandserviceswithouttheneedfortraditionalintermediarieslikebanksorfinancialinstitutions11.

DeFiapplicationsleveragesmartcontracts,whichareself-executingprogramsonblockchainnetworks,toautomateandenforcefinancialoperationssuchaslending,borrowing,trading,andinvesting.

AI.AIingeneraldefinesacollectionoftechnologiesenablingamachineorsystemtocomprehend,learn,act,reason,andsenselikeahuman12.AIsystemsleveragealgorithms,data,andcomputationalpowertoadaptandimproveovertime.

TheproliferationofAItoolsinrecentyearshasprovidedthefinancialindustrywiththepossibilitytointegrateitscapabilitiesintodiverseusecases.AIoffersnoteworthybenefits,includingenhancedoperationalefficiency,strengthenedregulatorycompliance,personalizedfinancialproducts,andadvanceddataanalytics

capabilities.FSRAlaunchedaninitiative,namedtheOpenRegulation(OpenReg),backin202213,aiming

9Section258ofFSMR(FINANCIALSERVICESANDMARKETSREGULATIONS)/rulebook/financial-services-and-markets

-regulations-2015-0

10Dixon,C.,2024.ReadWriteOwn:BuildingtheNextEraoftheInternet.RandomHouse.

11BIS(2023)|TheTechnologyofDecentralizedFinance(DeFi)

/publ/work1066.htm

12NationalProgramForArtificialIntelligence(2020)|AIGuide.ae/wp-content/uploads/2020/02/AIGuide_EN_v1-online.pdf

13ADGM(2022)|ADGM’sFinancialServicesRegulatoryAuthoritylaunchesitsAIinitiativeonOpenRegulation

/media/announcements/adgms

-financial-services-regulatory-authority-launches-its-ai-initiative-on-open-regulation

ADM

Academy

Researchcentre

ADGM

6

tomakeregulatorycontextmachine-readableregulations.TheprojectenablesRegTechcompaniesanddatasciencecommunitytousetheAItraininggroundtobuildthenextgenerationofAI-enabledRegTechsolutions.

Inthispaper,asapartofFSRA’songoingjourneytoembedAItechnologiesintoFSRAsupervisory

approaches,weelaborateontheAIadoptionforRegTechandSupTechforWeb3regulatoryactions/activities,takingintoconsiderationthevaluableinsightsprovidedinrecentreports14publishedbytheFinancialStabilityBoard(FSB),theregulatoryprinciplesoutlinedEUAIAct15,andtheriskframework

developedbyProjectMindForge16.

3–OPPORTUNITIESFORUTILISINGAIFORREGULATINGWEB3ACTIVITIES

TheregulatoryframeworkforWeb3hasnuancesthatdifferentiateitfromtraditionalregulationsduetotheuniquecharacteristicsofblockchaintechnology,smartcontracts,andspeedofWeb3innovation.

Globally,therecentfocusofregulatingWeb3hasbeenonVAsandtheirtradingplatforms.Thisincludes

enforcinganti-moneylaundering(AML)measuressuchasincorporatingKYT(knowyourtransactions)

solutionsandimplementingtravelrule17requirements,establishingprudentialguidelinesforstablecoinissuersand,morerecently,regulationofdecentralizedownerlessentitiessuchasDLTFoundationsanddecentralisedautonomousorganisations(DAOs).18Theseeffortstocreateregulatoryframeworksandimpositionof

safeguardstoprotectcustomersandinvestorsdemonstrateabroadeningacceptanceofVAandWeb3.

InexaminingtheinherentcharacteristicsofWeb3andVAforthefinancialregulator’sperspective,itiscrucialtoconsider(butnotlimitedto):

》Their24/7continuousoperationwithminimalhumanoversight,facilitatedbyself-executingsmartcontractsonDLTs;

》Theheightenedsecurityrisksduetovulnerabilitiesinsmartcontractcoding,potentialexploits,andrelianceondecentralizednetworks;and

》Theintroductionof‘new’conceptsthateitherrepurposeexistingtraditionalfinancialframeworkswithblockchaininnovationsorpresententirelynovelideaswithnohistoricalprecedent.

》ThedecentralizednatureofWeb3ensuresthattransactionsandsmartcontractsareimmutable,

enhancingtrustandtransparencybutmakingitchallengingtoaddresserrorslike“fat-finger”

mistakes,hacking,orunintendedoutcomes.

RegulatingWeb3activitiesintroducesseveralchallengesthatnecessitateinnovativeregulatoryapproachesandthedevelopmentofnewtoolstoenhancesupervisorymonitoringandenforcementcapabilities.

Nevertheless,thesechallengesalsopresentconsiderableopportunitiestoshapeabetterfuturefortheWeb3ecosystem.

14FSB(2024)|TheFinancialStabilityImplicationsofArtificialIntelligence

/2024/11/the-financial-stability-implications-of-artificial-intelligence/

15European

UnionArtificialIntelligenceActhttps://artificialintelligenceact.eu/

16MAS(2023)|MASPartnersIndustrytoDevelopGenerativeAIRiskFrameworkfortheFinancialSector

.sg/news/media-releases/2023/mas

-partners-industry-to-develop-generative-ai-risk-framework-for-the-financial-sector

17TheFATFRecommendations/en/publications/Fatfrecommendations/Fatf-recommendations.html

18ADGMDLTFoundationsFramework

/dlt-foundations

ADM

Academy

Researchcentre

ADGM

7

Fast-PacedInnovationandRiskIdenti?cation.ThenatureandfastpaceofinnovationinWeb3technologiesmakeitchallengingtopromptlyidentifyandmitigateemergingrisks.Thisdynamicenvironmentrequires

regulatoryprocessesandframeworkstoachieveagreatermeasureofresponsivenesstoensureregulatorsremainagileandabletoeffectivelyidentify,assess,andaddresspotentialrisks.

Thegapinresponsivenessincreasesthepotentialforfraudandmarketfailures.However,theseregulatory

challengesalsoopenupopportunitiestobuildframeworks“fromscratch”,allowingfortheintegrationof

forward-thinkingprinciplesthatcanadaptovertime.Thiscanencouragethedevelopmentofefficient

businessmodelstailoredtotheuniquecharacteristicsofWeb3,ultimatelyfosteringastableyetdynamic

marketplacethatalignswithbothregulatoryandindustrygrowthobjectives.AIcanplayaroleinfacilitatinginvestigationofsuchconcernsandintheconstructionofregulatoryframeworksbyrapidlyidentifying

enhancementsinregulatoryrulebookstoquicklyrespondtoWeb3developments.

AdvancedReal-TimeRiskMonitoring.EffectiveriskmonitoringintheWeb3ecosystemrequiresadvanced

toolscapableofreal-timeanalysisofextensiveblockchaindata.Giventhecontinuous24/7operation

ofDLTsandsmartcontracts,traditionalpoint-in-timeregulatoryapproachesoftenstruggletomanage

thevolumeandcomplexityofdatageneratedbytransactions.Consequently,thereisapressingneedfor

regulatorybodiestodevelopmoresophisticatedanalyticaltools.Implementingcontinuousmonitoring

systemsandautomatedriskmanagementtoolscanaidinmonitoringcompliancetoregulatoryrequirements,enablingproactiveresponsestopotentialthreats.

JurisdictionalComplexities.ThedecentralizednatureofWeb3activitiesoftencreatescross-jurisdictional

challengestoregulatoryapproaches.SinceeachregulatormayhaveadifferentapproachtogoverningVA,firmsmayfinditdifficultandcostlytomaintaincomplianceacrossmultiple,sometimesconflicting,regulatoryrequirementstherebyincreasingthetendencytopracticeregulatoryarbitrage.AI-drivenRegTechtoolscanpotentiallyhelpstreamlineandmanagethesecomplexitiesforfirms.Byautomatingroutinecompliance

tasks,identifyingoverlappingregulatoryrequirements,adaptingtonewrulesmoreefficiently,andassistingregulatoryreportingprocesses,AIcanreducecostsandoperationalburdens,ultimatelymakingiteasierforfirmstomeetdiverseregulatoryexpectations.

Inthefollowingsections,weconsidervariousscenarioswheretheuseofAIoffersbenefitstoregulatoryprocesses.

4–AIINNOVATIONS

TheevolutionofAItechnologieshasexperiencedsignificantadvancements,transformingoperational

andinnovationlandscapesacrossvariousindustries19IntheWeb3andVAspace,AIhasthepotentialtoextensivelyimproveregulatoryoversightandcomplianceefficiency.

ThissectionprovidesanoverviewofemergingAItechnologiesandhowAIinnovationscouldreshapethe

Web3regulatoryenvironment.ItstartswithabriefintroductionofwidelyusedAImodels(weonlyelaboratebrieflyofthepotentiallywidelyusedmodelsforregulatorydomain),followedbyusecasesinadoptingtheseAItechniquesforregulatoryactivities.WealsodiscussthekeychallengesofutilisingAIbeforeconsidering

possibledirectionsforfuturedevelopments.

19McKinsey(2024)|ThestateofAIinearly2024:GenAIadoptionspikesandstartstogeneratevalue

/capabilities/quantumblack/our-insights/

the-state-of-ai

ADM

Academy

Researchcentre

ADGM

8

4.1EmergingAITechnologies

MachineLearning(ML).MLisasubsetofAIthatspecializesinmakingpredictionsordecisionsbasedon

data.20MLalgorithmsexcelatanalysingvastarraysoftransactiondatatodetectpatternsandanomalies

indicativeoffraudulentactivitiesorcomplianceissues.Byemployingsupervised,unsupervised,and

reinforcementlearningtechniques,MLmodelscanadaptandimproveovertime,offeringregulatorspowerfultoolstoimprovemonitoringefficiencyandaccuracywithouttheneedforconstanthumanoversight.

NaturalLanguageProcessing(NLP).NLPfocusesonenablingcomputerstounderstandandprocesshumanlanguage(i.e.,text).21NLPcandeliverefficienciesforregulatoryreviewsandassessmentsthroughautomatedextractionandanalysisofkeyinformationfromvastamountsofdocumentsandcommunications.AdvancedNLPmodelshaveachievedsignificantprogressinbothunderstandingandgeneratinghuman-liketext,whichcanbeutilizedtoautomateresponsestobothregulatoryandpublicinquiries.22

However,NLPtechnologiesmayincludethepotentialformisinterpretationandbiasasNLPmodelsmaynotfullyaccountforvariedcontextortonedependingonculturalorsocietalnorms.Suchchallengesmayleadtoincorrectregulatoryresponsesoractionsifadoptedwithouthumanintervention.

GenerativeAI.GenerativeAIreferstoAItechnologiesthatcangeneratenewcontent(e.g.,text,images,andothermedia)basedonexistingdata.23

However,NLPtechnologiesmayincludethepotentialformisinterpretationandbiasasNLPmodelsmaynotfullyaccountforvariedcontextortonedependingonculturalorsocietalnorms.Suchchallengesmayleadtoincorrectregulatoryresponsesoractionsifadoptedwithouthumanintervention.

AIAgents.AIagentsarespecialisedgenerativeAImodelimplementationsthatcanperformcomplextaskswithprogrammedworkflows,suchasautomatingcustomerserviceinteractions,generatinglegaland

regulatorydocuments,orevenconductingvirtualnegotiationsonbehalfofhumanoperators.24

Intheregulatorycontext,generativeAIandAIagentshavemanypotentialapplications.Forinstance,

theycanbeusedbyregulatedentitiestoautomatethegenerationofdetailedperiodicoron-demand

compliancereports.SuchAItechnologiescanalsobeemployedbyregulatorstoanalyselargedatasetsofregulatoryreturnsandgenerateashortlistofpotentialnon-complianceandriskindicators.However,akintothelimitationsinherentinNLPtechnologies,thecurrentavailablegenerativeAImodels,whicharelargelybasedonlargelanguagemodels(LLMs),havelimitationsinaccuracyandreliabilityofoutputsduetothe

potentialforhallucinationandcontextualmisinterpretation.

GeneralAI.GeneralAIreferstohighlyautonomoussystemscapableofperforminganycognitivetaskthat

ahumanbeingcanundertake.25UnlikegenerativeAI,whichisdesignedforspecificcontentcreationtasks,

generalAIischaracterizedbyitsversatilityandabilitytoadapttoawiderangeofscenarioswithoutprior

specificprogramming.Whilestillconceptual,generalAIcouldfacilitatehighlyadaptivesystemsforregulatoryoversightandcompliancemanagementautonomouslyadjustingtonewregulationsandthecomplexitiesoflegalcompliancewithlittletonohumanintervention.

20IBM(2023)|AIvs.machinelearningvs.deeplearningvs.neuralnetworks:What’sthedifference?

/think/topics/ai-vs-machine-learning-vs-deep-learning-vs-neural-networks

21IBM

(2024)|whatisNLP?/topics/natural-language-processing

22PwC(2022)|UnderstandingalgorithmicbiasandhowtobuildtrustinAI/us/en/tech-effect/ai-analytics/algorithmic-bias-and-trust-in-ai.html

23McKinsey(2024)|WhatisgenerativeAI?/featured-insights/mckinsey-explainers/what-is-generative-ai

24

/index/introducing-gpts

25

/2024/04/05/what-is-agi-artificial-general-intelligence

ADM

Academy

Researchcentre

ADGM

9

4.2AISolutionsintheWeb3RegulatoryDomain

InthissectionweexplorehowdiversetypesofAItechnologiesmaybeimplementedintheWeb3regulatoryspacetomeetthechallengesofmonitoring,enforcement,andcompliancemanagement.Wesplitthese

technologiesintotwomajorcategories:applicationsusingnarrowAIandthoseusinggenerativeAI.Note

thatnarrowAIreferstoartificialintelligencesystemsthataredesignedtoperformspecifictasksandoperateunderlimitedconstraints.26Theyarealsoreferredtoas“specificAI”or“weakAI”.

4.2.1NarrowAI-basedApplications

RegulatoryReportingTools.AI-drivenregulatoryreportingtoolscanautomatethecollection,submission

andanalysisofregulatoryreturnsandattestationreports.2728Thesesystemsutilizeadvanceddatamining

andprocessingalgorithmstoextractandorganizeinformationfromvastdatasetstofacilitateseamless

regulatoryreporting.Beyondreportingautomation,AItoolsthatperformpredictiveanalyticsmayenable

regulatedentitiestoidentifyriskfactorsthatcanreducethepotentialforcompliancefailures.Forexample,AIcanbeemployedtomonitorandpredictfinancialrisksthatmayhindercompliancewithliquidityandcapitalobligations.

RiskPro?ling.AIsystemsdedicatedtoriskprofiling29cananalyseandcategorizevirtualassetsorfinancial

entitiesbasedontheirriskcharacteristicsandapplicableregulatoryrequirements.Thesesystemscan

evaluatehistoricalperformance,marketbehaviour,andexternalfactorstomaintainadynamicriskprofile.Bycontinuouslylearningfromnewdataandregulatoryupdates,theseAIprofilerscankeepprofilesapacewiththeevolvingfinanciallandscape30.

KnowYourTransaction(KYT).Utilizinggraphanalyticsandgraphneuralnetworks(GNNs),AI-drivenKYT

andanomalydetectionsystemscanbespecificallydesignedtomonitorandanalyseaccounts31and

transactionsonblockchainnetworks.ByleveragingonthecapabilitiesofAItoexaminecomplexblockchaintransact