版權說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權,請進行舉報或認領
文檔簡介
AIGovernanceinPracticeReport2024
Tableofcontents
What'sinside?
Executivesummary 3
PartI.UnderstandingAIandgovernance 6
PartII.Thedatachallenge 15
PartIII.Theprivacyanddataprotectionchallenge 23
PartIV.Thetransparency,explainability
andinterpretabilitychallenge 32
PartV.Thebias,discriminationandfairnesschallenge 41
PartVI.Thesecurityandrobustnesschallenge 50
PartVII.AIsafety 55
PartVIII.Thecopyrightchallenge 61
PartIX.Third-partyAIassurance 65
Conclusion 69
Contacts 70
AIGovernanceinPracticeReport2024|2
Executive
summary
Recentandrapidlyadvancing
breakthroughsinmachine
learningtechnologyhaveforevertransformedthelandscapeofAI.
AIsystemshavebecomepowerfulenginescapableof
autonomouslearningacrossvastswathsofinformationand
generatingentirelynewdata.Asaresult,societyisinthemidstofsignificantdisruptionwiththesurgeinAIsophisticationandtheemergenceofaneweraoftechnologicalinnovation.
AsbusinessesgrapplewithafutureinwhichtheboundariesofAIonlycontinuetoexpand,theirleadersfacetheresponsibilityofmanagingthevariousrisksandharmsofAI,soitsbenefitscanberealizedinasafeandresponsiblemanner.
Critically,thesebenefitsareaccompaniedbyserious
considerationsandconcernsaboutthesafetyofthistechnologyandthepotentialforittodisrupttheworldandnegatively
impactindividualswhenleftunchecked.Confusionabouthowthetechnologyworks,theintroductionandproliferationofbiasinalgorithms,disseminationofmisinformation,andprivacy
rightsviolationsrepresentonlyasliverofthepotentialrisks.
Thepracticeof
AIgovernance
isdesignedtotacklethese
issues.Itencompassesthegrowingcombinationofprinciples,laws,policies,processes,standards,frameworks,industry
bestpracticesandothertoolsincorporatedacrossthedesign,development,deploymentanduseofAI.
AIGovernanceinPracticeReport2024|3
→Executivesummary
TABLEOFCONTENTS↑
iapp
GlobalAIprivateinvestment
(USDbillion,2021)
94
47
42
38
22
18
10
8
4
201320142015201620172018201920202021
Whilerelativelynew,thefieldofAIgovernanceismaturing,withgovernmentauthorities
aroundtheworldbeginningtodeveloptargetedregulatoryrequirementsandgovernance
expertssupportingthecreationofacceptedprinciples,suchastheOrganisationfor
EconomicCo-OperationandDevelopment's
AI
Principles
,emergingbestpracticesandtoolsforvarioususesofAIindifferentdomains.
Therearemanychallengesandpotential
solutionsforAIgovernance,eachwithuniqueproximityandsignificancebasedonan
organization'srole,footprint,broaderrisk-governanceprofileandmaturity.Thisreportaimstoinformthegrowing,increasingly
empoweredandincreasinglyimportant
communityofAIgovernanceprofessionalsaboutthemostcommonandsignificant
challengestobeawareofwhenbuilding
andmaturinganAIgovernanceprogram.
Itoffersactionable,real-worldinsights
intoapplicablelawandpolicy,avarietyofgovernanceapproaches,andtoolsusedto
managerisk.Indeed,someofthechallengestoAIgovernanceoverlapandrunthrough
arangeofthemes.Therefore,anemergingsolutionforonethematicchallengemay
alsobeleveragedforanother.Conversely,in
certaincircumstances,specificchallengesandassociatedsolutionsmayconflictandrequire
reconciliation
withotherapproaches.Someofthesepotentialoverlapsandconflictshavebeenidentifiedthroughoutthereport.
TECHNOLOGY
FTI"
CONSULTING
AIGovernanceinPracticeReport2024|4
Questionsaboutwhetherandwhen
organizationsshouldprioritize
AIgovernancearebeinganswered:
"yes"and"now,"respectively.
FTI"
CONSULTING
TECHNOLOGY
iapp
Questionsaboutwhetherandwhenorganizationsshouldprioritize
AIgovernancearebeinganswered:"yes"and"
now
,"respectively.
Thisreportis,therefore,focusedonhoworganizationscanapproach,buildandleverageAIgovernanceinthecontextoftheincreasingly
voluminousandcomplexapplicablelandscape.
UzmaChaudhry
IAPPAIGovernance
CenterResearchFellow
AshleyCasovan
IAPPAIGovernanceCenterManagingDirector
JoeJones
IAPPDirectorofResearchandInsights
LuisaResmerita
FTITechnologySeniorDirector
NinaBryant
FTITechnologySeniorManagingDirector
MichaelSpadea
FTITechnologySeniorManagingDirector
AIGovernanceinPracticeReport2024|5
PartI.
UnderstandingAIand
governance
ComponentsofanAIsystemandtheirgovernance
TounderstandhowtogovernanAIsystem,itisimportanttofirst
understandwhatanAIsystemis.TheEUAIAct,forexample,
definesanAIsystemas"amachine-basedsystemthatisdesigned
tooperatewithvaryinglevelsofautonomyandthatmayexhibit
adaptivenessafterdeployment,andthat,forexplicitorimplicit
objectives,infers,fromtheinputitreceives,howtogenerateoutputssuchaspredictions,content,recommendations,ordecisionsthat
caninfluencephysicalorvirtualenvironments."
AsindicatedintheOECD'sFrameworkfortheClassificationofAI
systems,AIsystemsarecomprisedofdatausedtotrainandoperateasystem,model,outputandcontext.WhileamodelisafundamentalbuildingblockofanAIsystem,asinglemodelseldomoperatesin
isolation.Instead,multipleAImodelscometogetherandinteract
witheachothertoformcomplexAIsystems.Additionally,AIsystemsareoftendesignedtointeractwithothersystemsforsharingdata,
facilitatingseamlessintegrationintoreal-worldenvironments.ThisresultsinanetworkofAIsystems,eachwithitsspecializedmodels,workingtogethertoachievealargergoal.
WithAIpoisedtorevolutionisemanyaspectsofourlives,freshcooperativegovernanceapproachesareessential.Effectivecollaborationbetweenregulatoryportfolios,withinnationsaswellasacrossborders,iscrucial:bothtosafeguardpeoplefromharmandtofosterinnovationandgrowth.
KateJones
U.K.DigitalRegulationCooperationForumCEO
AIGovernanceinPracticeReport2024|6
→PartI.UnderstandingAIandgovernance
TABLEOFCONTENTS↑
iapp
NavigatingAIgovernancesources
Giventhecomplexityandtransformative
natureofAI,significantworkhasbeendonebylawandpolicymakersonwhatisnowa
vastandgrowingbodyofprinciples,laws,
policies,frameworks,declarations,voluntarycommitments,standardsandemergingbestpracticesthatcanbechallengingtonavigate.Manyofthesevarioussourcesinteractwitheachother,eitherdirectlyorbyvirtueoftheissuescovered.
AIprinciples,suchastheOECD'sAIPrinciplesorUNESCO'sRecommendationontheEthics
ofAI,canshapeglobalstandards,especiallywhennationalgovernmentspledgeto
voluntarilyincorporatesuchguidanceinto
theirdomesticAIgovernanceinitiatives.
Theyprovideanonbinding,principled
approachtoguidelegal,policyandindustry
effortstowardtacklingthematicchallenges.AlgorithmWatchcreatedaninventoryoftheseprinciples,identifying167
reports
.
Lawsandregulationsincludeexistinglegislationthatisnotspecificbutis
nonethelessapplicabletoAI,aswellas
emerginglegislationthatmorespecifically
addressesthegovernanceofAIsystems,such
astheEUAIAct.TheEUAIActistheworld's
firstcomprehensiveAIregulation.Although
jurisdictionalvariationscanbeobservedacross
TECHNOLOGY
FTI"
CONSULTING
theemergingglobalAIregulatorylandscape,manydraftregulationsadoptarisk-based
approachsimilartotheEUAIAct.
TheEUAIActmandatesAIgovernancestandardsbasedontheriskclassificationofAIsystemsandtheorganization'sroleasanAIactor.Certain
AIsystemsaredeemedtoposeunacceptable
riskandareprohibitedbylaw,subjecttovery
narrowexceptions.Thebulkoftherequirementsimposedbytheactapplytoprovidersofhigh-riskAIsystems,althoughdeployersandresellers,
namelydistributersandimporters,arearealsosubjecttodirectobligations.
Theactimposesregulatoryobligationsat
enterprise,productandoperationallevels,
suchasestablishingappropriateaccountabilitystructures,assessingsystemimpact,providingtechnicaldocumentation,establishingrisk
managementprotocolsandmonitoring
performance,amongotherkeyrequirements.Inthecontextofthegrowingvarietyof
generativeAIusecasesandadoptionof
solutionsembeddinggenerativeAIsuchasMSCopilot,generalpurposeAI-specificprovisionsareanothercrucialcomponentoftheEUAI
Act.Dependingontheircapabilities,reachandcomputingpower,certainGPAIsystemsare
consideredtopresentsystemicriskandattractbroadlysimilarobligationstothoseapplicabletohigh-riskAIsystems.
AIgovernanceisabouttogetalot
harder.Theinternalcomplexityof
governingAIisgrowingasmore
internalteamsadoptAI,newAI
featuresarebuilt,andthesystems
getcomplex,butatthesametime,
theexternalcomplexityisalsosetto
growrapidlywithnewregulations,
customerdemands,andsafety
researchevolving.
Theorganizationswhohaveinvested
instructuredAIgovernancealready
havealegupandwillcontinueto
haveacompetitiveadvantage.
AndrewGamino-Cheong
TrustibleAICo-founderandChiefTechnologyOfficer
AIGovernanceinPracticeReport2024|7
Inadditiontobindinglegislation,voluntaryAIframeworks,suchastheNationalInstitute
ofStandardsandTechnology'sAIRisk
IAPPGlobalAILawandPolicyTracker
ManagementFrameworkandtheInternationalOrganizationforStandardization'sAIStandards,
offerstructuredandactionableguidance
stakeholderscanelecttousetosupport
theirworkonimplementingAIgovernance.
Voluntarycommitmentsareoftendevelopedtobringdifferentstakeholdersclosertoasharedunderstandingofidentifying,assessingand
managingrisks.Standardsserveasbenchmarks
thatcandemonstratecompliancewithregulatoryrequirements.
Internationaldeclarationsandcommitments
memorializesharedcommitments,oftenbetweengovernments,tospecificaspectsorbroad
swathesofAIgovernance.Whilenotbinding,
suchcommitmentscan,ataminimum,indicateacountry'ssupportforandintentiontoadvanceAIgovernanceinparticularorgeneralways,evenatthehighestoflevels.
NavigatingagrowingbodyofdraftAIlaws,
regulations,standardsandframeworkscan
bechallengingfororganizationspioneering
Thismapshowsthejurisdictionsinfocusandcoveredbythe
IAPPGlobalAILawandPolicyTracker
.ItdoesnotrepresenttheextenttowhichjurisdictionsaroundtheworldareactiveonAIgovernancelegislation.TrackerlastupdatedJanuary2024.
withAI.ByunderstandingtheiruniqueAIriskprofileandadoptingarisk-basedapproach,
organizationscanbuildarobustandscalableAIgovernanceframeworkthatcanbedeployedacrossjurisdictions.
AIGovernanceinPracticeReport2024|8
iapp
Thefollowingare
examples
ofsomeofthemostprominentandconsequential
AIgovernanceefforts
:
Principles
→OECDAIPrinciples
→EuropeanCommission'sEthicsGuidelinesforTrustworthyAI
→UNESCORecommendationontheEthicsofAI
→TheWhiteHouseBlueprintforanAIBillofRights→G7HiroshimaPrinciples
→EUAIAct
Lawsand
→EUProductLiabilityDirective,proposed→EUGeneralDataProtectionRegulation→Canada–AIandDataAct,proposed
→U.S.AIExecutiveOrder14110
regulations
→SectoralU.S.legislationforemployment,housingandconsumerfinance→U.S.statelaws,suchasColoradoAIAct,SenateBill24-205
→China'sInterimMeasuresfortheManagementofGenerativeAIServices
→TheUnitedArabEmiratesAmendmenttoRegulation10toincludenewrulesonProcessingPersonalDatathroughAutonomousandSemi-autonomousSystems
→DigitalIndiaAct
AIframeworks
→OECDFrameworkfortheclassificationofAISystems→NISTAIRMF
→NISTSpecialPublication1270:TowardsaStandardforIdentifyingandManagingBiasinAI→SingaporeAIVerify
→TheCouncilofEurope'sHumanRights,Democracy,andtheRuleofLawAssuranceFrameworkforAIsystems
Declarations
→BletchleyDeclaration
andvoluntary
→TheBiden-HarrisAdministration'svoluntarycommitmentsfromleadingAIcompanies
commitments
→Canada'sguideontheuseofgenerativeAI
→ISO/IECJTC1SC42
Standardsefforts
→TheInstituteofElectricalandElectronicsEngineersStandardsAssociationP7000
→TheEuropeanCommitteeforElectrotechnicalStandardization
AIstandardsforEUAIAct
→The
VDEAssociation'sAIQualityandTestingHub
→TheBritishStandardsInstitutionand
AlanTuringInstituteAIStandardsHub
→Canada's
AIandDataStandardsCollaborative
TECHNOLOGY
FTI"
CONSULTING
Itisimportanttotakean
ecosystemapproachtoAI
governance.Policymakersand
industryneedtoworktogether
atplatformssuchastheAIVerify
Foundationtomakesenseofthe
opportunitiesandrisksthatthis
technologybrings.Theaimisto
findcommonguardrailstomanage
keyrisksinordertocreatea
trustedecosystemthatpromotes
maximalinnovation.
DeniseWong
SingaporeInfocommMediaDevelopmentAuthorityAssistantChiefExecutive,DataInnovation&ProtectionGroup
AIGovernanceinPracticeReport2024|9
AIRisks
REPUTATIONAL
Riskof
damageto
reputation
andmarket
competitiveness
INDIVIDUALSANDSOCIETY
Riskofbias
orother
detrimental
impacton
individuals
LEGALANDREGULATORY
Riskof
noncompliance
withlegaland
contractual
obligations
FINANCIAL
Riskoffinancial
implications,
e.g.,fines,legal
oroperational
costs,or
lostprofit
TheAIgovernanceimperative
Withprivateinvestment,globaladoptionratesandregulatoryactivityontherise,aswellas
thegrowingmaturityofthetechnology,AIis
increasinglybecomingastrategicpriorityfororganizationsandgovernmentsworldwide.
Organizationsofallsizesandindustriesare
increasinglyengagingwithAIsystemsatvariousstagesofthetechnologyproductsupplychain.
Theexceptionaldependenceonhighvolumes
ofdataandendlesspracticalapplicabilitythat
makeAItechnologyadisruptiveopportunity
canalsogenerateuniquelymultifacetedrisksforbusinessesandindividuals.Theseincludelegal,regulatory,reputationaland/orfinancialriskstoorganizations,butalsoriskstoindividualsandthewidersociety.
iapp
FTI"
CONSULTING
TECHNOLOGY
AIGovernanceinPracticeReport2024|10
iapp
Enterprisegovernance
AIgovernancestartswithdefiningthecorporatestrategyforAIbydocumenting:
→TargetoperatingmodelstosetoutclearrolesandresponsibilitiesforAIrisk.
→Complianceassessmentstoestablish
programmaturityandremediationpriorities.
→Accountabilityprocessestorecordanddemonstratecompliance.
→Policiesandprocedurestoformulatepolicystandardsandoperationalprocedures.
→Horizonscanningtoenhanceandaligntheprogramwithongoingregulatorydevelopments.
Productgovernance
AIgovernancealsorequiresenterprisepolicystandardstobeappliedattheproductlevel.OrganizationscanensuretheirAIproductsmatchtheirenterprisestrategybyusing:
TECHNOLOGY
FTI"
CONSULTING
→Systemimpactassessmentstoidentifyandaddressriskpriortoproductdevelopmentordeployment.
→Qualitymanagementprocedurestailoredtothesoftwaredevelopmentlifecycletoaddressriskbydesign.
→Riskandcontrolsframeworkstodefine
AIriskandtreatmentbasedonwidely
recognisedstandardssuchasISOandNIST.
→Conformityassessmentsanddeclarationstodemonstratetheirproductsarecompliant.
→Technicaldocumentationincludingstandardizedinstructionsofuseandtechnicalproductspecifications.
→Post-marketmonitoringplanstomonitor
productcompliancefollowingmarketlaunch.
→Third-partyduediligenceassessmentstoidentifypossibleexternalriskandinformselection.
AIGovernanceinPracticeReport2024|11
FTI"
CONSULTING
TECHNOLOGY
iapp
Operationalgovernance
Theorganization'sAIstrategymustultimatelybeoperationalizedthroughoutthebusinessthroughthedevelopmentof:
→Performancemonitoringprotocolstoensuresystemsperformadequatelyfortheirintendedpurposes.
→Transparencyandhumanoversightinitiativestoensure
individualsareawareandcanmakeinformedchoiceswhentheyinteractwithAIsystemsorwhenAI-powereddecisionsaremade.
→Incidentmanagementplanstoidentify,escalateandrespondtoseriousincidents,malfunctionsandnationalrisksimpactingAIsystemsandtheiroperation.
→Communicationstrategiestoensuretransparencytowardinternalandexternalstakeholdersinrelationtothe
organization'sAIpractices.
→TrainingandawarenessprogramstoenablestaffwithrolesandresponsibilitiesforAIgovernancetohelpthemunderstandandperformtheirrespectiveroles.
→Skillsandcapabilitiesdevelopmenttoassesshumanresourcescapabilitiesandreviewordesignjobrequirements.
AIGovernanceinPracticeReport2024|12
iapp
TheAIlifecycle
PLANNING
→Plananddocumentthesystem'sconceptandobjectives.
→Planforlegaland
regulatorycompliance.
DESIGN
→Gatherdataandcheckfordataquality.
→Documentand
assessmetadataandcharacteristicsof
thedataset.
→Considerlegaland
regulatoryrequirements.
DEVELOPMENT
→Selectthealgorithm.→Trainthemodel.
DEPLOYMENT
→Pilotandperform
compatibilitychecks.
→Verifylegaland
regulatorycompliance.
→Monitorperformanceandmitigateriskspostdeployment.
→Carryouttesting,
validationand
verification.
→Calibrate.
→Carryoutoutput
interpretation.
UnderstandingthatAIsystems,likeall
products,followalifecycleisimportantastherearegovernanceconsiderationsacrossthelifecycle.The
NISTAIRMF
setsouta
comprehensivearticulationoftheAIsystem
lifecycleandincludesconsiderationsfor
testing,evaluation,validation,verification
andkeystakeholdersforeachphase.Amoresimplifiedsamplelifecycleisincludedabove,alongwithsometop-levelconsiderations.
TECHNOLOGY
FTI"
CONSULTING
AneffectiveAIgovernance
modelisaboutcollective
responsibilityandcollective
businessresponsibility,
whichshouldencompass
oversightmechanismssuch
asprivacy,accountability,
compliance,amongothers.
Thisresponsibilityshouldbe
sharedbyeverystakeholder
whoispartoftheAI
governancechain.
VishalParmar
BritishAirwaysGlobalLeadPrivacyCounselandDataProtectionOfficer
AIGovernanceinPracticeReport2024|13
→PartI.UnderstandingAIandgovernance
TABLEOFCONTENTS↑
→HOWTO
Navigatedevelopersfromdeployers
Variouscontractualandregulatoryobligationsmayarise
dependingonwhetheranorganizationisavendororbuyer,
orifitsourcesexternalservicessuchashardware,cloudordatacollection,forthedevelopmentandoperationsofitsAIsystem.
PriorIAPP
research
foundmorethan70%oforganizationsrelyatleastsomewhatonthird-partyAI,sotheresponsibilityforensuringtheAIsystemissafeandresponsiblemaybe
spreadacrossmultipleroles.
→TheWorldEconomicForumputtogetherauseful
toolkit
tohelpthosewhoareprocuringAIsystems.
Inboth
currentlegislation
and
proposedlegislation
wearestartingtoseedifferentobligationsforthosewhoprovideandsupplyAIversusthosewhodeployAI.Understandingwhetheryouareadeveloperand/ordeployeris
importanttoensuringyoumeetcomplianceobligations.Oncethisisunderstood,itispossibletoestablishAI-governanceprocessesfor
procurement
,includingevaluationsandcontractstoavoidtakingonadditionalliabilities.
iapp
FTI"
CONSULTING
TECHNOLOGY
AIGovernanceinPracticeReport2024|14
PartII.Thedatachallenge
DataisanintegralpartoftrainingandoperatinganAIsystem.
MostAIrequiressizeableamountsofhigh-qualitydata,
especiallyduringthetrainingphasetomaximizethemodel'sperformance,aswellastoensurethedesiredandaccurate
output.WiththeadvancementofnewAItechnologies,modelsarerequiringincreasinglymoredata,whichmaycomefromavarietyofsources.GiventheimportanceofthedatausedtofueltheAIsystem,itisimportanttounderstandwhatdataisbeingused;how,whereandbywhomitwascollected;from
whomitwascollected;ifitistherightdataforthedesired
outcome;andhowitwillbemanagedthroughoutthelifecycle.
Accessingdataandidentifyingdatasources
Understandingwheredatacomesfromandhowitiscollectedis
notonlynecessaryforAIsystems,butalsoforbuildingtrustinAIbyensuringthelawfulnessofdatacollectionandprocessing.Suchdocumentationcanassistwith
datatransparency
andimprovetheAIsystem'sauditabilityaswell.
Althoughdatamayoriginatefrommultiplesources,itcanbebroadlycategorizedintothreetypes:first-partydata,publicdataandthird-partydata.
AIGovernanceinPracticeReport2024|15
→PartII.Thedatachallenge
TABLEOFCONTENTS↑
iapp
First-partydata
Thisreferstodatacollecteddirectlyfrom
individualsbyanorganizationthroughtheirowninteractionsandtransactions.Suchdatamay
originatefromsourcessuchaswebsitevisits,
customerfeedbackandsurveys,subscriptions,andcustomerrelationshipmanagementsystems,amongothers.Thisdataisextremelyvaluablefororganizationsasitprovidesdirectandfirsthandinsightsintoindividuals'behavior.
First-partydatacanbecollectedfromvarious
sources.Identifyingthedatachannelsand
documentingthesourcewillnotonlyhelpthe
organizationdeterminewhattypesofdata,e.g.,text,numerical,imageoraudio,willbecollectedfromeachsource,butalsoalertthelegalteam
aboutwherelegalcompliancewillberequiredontheorganization'spart.
Publicdata
Thisreferstodatathatisavailabletothewider
publicandencompassesarangeofsources,
suchaspubliclyavailablegovernmentrecords,publications,andopensourceandweb-scrapeddata.Publicdataisavaluableresourcefor
researchersandinnovatorsasitprovidesreadilyavailableinformation.Publicdatacancome
frommultiplesources.
TECHNOLOGY
FTI"
CONSULTING
Whileitisarduousandcumbersometo
maintaindatalineageforpublicdatasets,
itisimportantforupholdingorganizational
reputationandfosteringusertrust,legal
complianceandAIsafetyoverall.Alackof
understandingofwheredatacomesfrom
eventuallyleadstoalackofunderstandingofthetrainingdatasetandmodelperformance,whichcanreinforcetheblack-boxproblem.
Therefore,intheinterestoftransparency,
trackinganddocumentingpublic-datasourcesasmuchaspossiblemayprovebeneficialfortheorganization,asitcanlatersupportothertransparencyefforts,suchasdrawingupdata,modelorsystemcards.
Moreover,withoutknowledgeofpublic-datasources,theorganizationmayinadvertentlytraintheAIsystemonpersonal,sensitive
orproprietarydata.Fromtheprivacy
standpoint,thiscanbeproblematicincasesof
dataleakage
,wherepersonallyidentifiabledatamaybeexposed.AIsecuritychallengesmayalsobeamplifiedifdatawasprocured
fromunsafepublicsources,asthatcarriestheriskofintroducingmaliciousbugsintothesystem.Itmayalsoleadtobiasesin
theAIsystem.
Ethicaldevelopment
practicesstartwith
responsibledata
acquisitionand
managementsystems,
aswellasreview
processesthattrack
thelineageof
sourceddata.
ChristinaMontgomery
IBMVicePresidentand
ChiefPrivacyandTrustOfficer
AIGovernanceinPracticeReport2024|16
Withoutunderstandingthequality
ofthedatabeingingestedintoan
AImodel,youmaynotknowthe
qualityoftheoutput.Companies
mustestablishanddefinewhat‘data
quality’involvesandconsistsof,asthis
determinationishighlycontextualfor
anyorganization,andcandependon
businessgoals,usecases,focusareas
andfitnessforpurpose.
Regardlessofcontext,thereare
minimumbaselineattributeswhichcan
andshouldbeestablished:accuracy,
completeness,consistencyand
validity.Timelinessanduniqueness
mayalsobeimportanttoestablishing
fitnessforpurpose.
DeraNevin
FTITechnologyManagingDirector
FTI"
CONSULTING
TECHNOLOGY
iapp
Anorganizationcanbeginbyestablishinga
clearunderstandingofhowandwhypublic
dataisbeingcollected,howitalignswiththe
purposestheAIsystemwillfulfil,ifandhow
systemaccuracywillbeaffectedbyusing
publicdata,whatthetrustworthysourcesfor
gatheringpublicdataare,iftheorganization
hasrightstousethepublicdata,andotherlegalconsiderationsthatmayhavetobetakeninto
account,particularlygiventhatpublicdataistreateddifferentlyacrossjurisdic
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
- 4. 未經(jīng)權益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負責。
- 6. 下載文件中如有侵權或不適當內(nèi)容,請與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。
最新文檔
- 浙江安防職業(yè)技術學院《中學語文課程教學論》2023-2024學年第一學期期末試卷
- 數(shù)獨答案生成器-20220303010135
- 數(shù)字教學實踐報告
- 經(jīng)典搞笑語錄集錦
- 部編版2024-2025學年六年級上語文寒假作業(yè)(七)(有答案)
- 浙江省杭州市拱墅區(qū)源清中學2024-2025學年高一(上)期中物理試卷(含答案)
- 2025屆吉林省前郭爾羅斯蒙古族自治縣第五高級中學高三上學期第五次考試歷史試題(綱要上下 選擇性必修三冊)(含答案解析)
- 《色達佛學院全景》課件
- 伊春職業(yè)學院《臺詞與表演》2023-2024學年第一學期期末試卷
- 二零二五年度城市綜合體土石方運輸及配套設施合同3篇
- ppr管件注塑工藝
- 職業(yè)技能鑒定要素細目表和卷庫編寫要求
- 公司設備轉讓合同協(xié)議書
- 2023年全國統(tǒng)一建筑工程預算工程量計算規(guī)則完整版
- 教科版四年級科學下冊第三單元巖石與土壤4.制作巖石和礦物標本(教學設計)教案
- 大學《工程力學》期末考試試題庫含詳細答案
- 2022年湖北省武漢市中考數(shù)學試卷含解析
- TLFSA 003-2020 危害分析與關鍵控制點(HACCP)體系調味面制品生產(chǎn)企業(yè)要求
- LY/T 2244.3-2014自然保護區(qū)保護成效評估技術導則第3部分:景觀保護
- 紀律教育月批評與自我批評五篇
- GB/T 26480-2011閥門的檢驗和試驗
評論
0/150
提交評論