CENTERFORDATAINNOVATION

1

ExploringData-SharingModelstoMaximizeBenefitsFromData

ByGillianDiebold|October16,2023

Data-driveninnovationhasthepotentialtobeamassiveforceforprogress.Datasharingenablesorganizationstoincreasetheutilityandvalueofthedatatheycontrolandgainaccesstoadditionaldatacontrolledbyothers.Thisreportevaluatestheadvantagesanddisadvantagesofsixcommondata-sharingmodelsandoffers

recommendationsforpolicymakerstopromotegreateruptakeofthesedata-sharingmodelstomaximizethe

economicandsocialbenefitsofdataintheUnitedStates.

Individualsandorganizationsusedatatomakebetterdecisionsandobtainbetterinsights,leadingtobenefitsinabroadrangeofareas.

1

Buttouse

datatoitsfullestpotential,individualsandorganizationsneedtobeabletocombine,augment,andanalyzeinformationfromdifferentsources.Intheprivatesector,datasharingenablesbusinessestoinnovatewithpartners,suchastacklingcommonchallengesandprovidingbetterexperiencestoconsumers.Inthepublicsector,itenablesgovernmentagenciestobuild

oninformationcollectedbyotherorganizationstomakebetterdecisions,offerpersonalizedservices,engageinevidence-basedpolicymaking,andgleannewinsights.Andamongacademicsandnonprofitorganizations,datasharingadvancesscientificbreakthroughsandenablesdatafor

socialgood.

Butattainingthesebenefitsrequiresenablingdatasharingtoitsfullestpotentialsothatthosewhocanusedataproductivelyhaveaccesstoit.Unlikemostresources,suchaslandoroil,dataisnonrivalrous,meaningthesupplyofdataisnotreducedwhenothersuseit.Datacanbeusedmultipletimesandinmultiplewaysbyvariousentitieswithoutbeing

depleted.

2

WhilemanyorganizationsintheUnitedStatessharedataincertaininstances,manyoftheseinitiativesareadhoc,andtherearefewbestpracticesforsharingdata.IfpolicymakerswanttheUnitedStatesto

CENTERFORDATAINNOVATION

2

havearobustAIanddata-drivensociety,theyneedtotakestepstoboostdatasharing.

Thisreportoffersastepinthatdirectionbyevaluatingthebenefitsanddrawbacksofsixdifferentdata-sharingmodelsandoffers

recommendationsonhowpolicymakerscanimplementorexpandtheuseofcertainmodels.Giventhatdifferentmodelsservedifferentneeds,

policymakersdonotneedtopickaone-size-fits-allsolution,butrather

shouldfacilitatetheadoptionofmultipledata-sharingmechanismsintheirpursuitofadata-drivensociety.

DATA-SHARINGMODELS

Datasharingistheprocessofmakingdataaccessibletoothers,whetheritbebetweenorwithinorganizationsorbetweenindividualsand

organizations.Approachestodatasharingcanvarywidelyandcaninvolvevarioustypesofactorsandhavedifferinggoals.Forexample,two

businessesmayusecontractualagreementstosharedatatofacilitatecollaborationonalarge-scaleproject.Ormultipleindividualsmaysharedatawiththroughanindependentorganizationforfinancialgain.

Data-sharingmodelslargelyvarybasedonwhocontributesthedata,whohasaccesstothedata,whostoresandmanagesthedata,andwho

benefitsfromthedatasharing.Datacontributorscanbeanyactorthat

ownsorcreatesdata,includingindividuals,privatecompanies,governmentagencies,nonprofits,andresearchinstitutions.Likewise,thosesame

actorscanalsobetheonesreceivingdatainadata-sharingarrangement.

Forexample,agovernmentagencymightsharehealthdatawith

pharmaceuticalcompaniesinvestigatingnewdrugs,orapharmaceuticalcompanymightshareitsdataonvaccinedistributionwiththegovernmentorpublichealthresearchers.Theseagreementscanbeone-way,where

oneactorsharesdatawithanotherpartyinordertoreceivespecific

insights,orreciprocal,whereeachactorreceivesdata.Lastly,data-sharingmodelsdifferdependingonwhoreceivesandstoresthedata,suchasthedataowneroranintermediaryinstitution.Thesefactorsarethecore

differencesamongdata-sharingmodels.

Thefollowingsectionexploresandevaluatessixdifferentmodels,

illustratingtheirrespectivestrengthsandweaknessesandoffering

recommendationsforU.S.policymakersonhowtobestimplementandincreasedatasharingacrossthenation.

Data-SharingPartnerships

Data-sharingpartnershipsinvolvecollaborativeeffortsbetweendifferententities,suchasacademicinstitutions,researchorganizations,industrypartners,individualconsumers,andgovernmentagencies,toshareandexchangedataforthepurposeofconductingresearch,collaboratingonnewproducts,andenhancingevidence-baseddecision-making.These

partnershipsaimtoleveragethecollectiveexpertise,resources,anddata

CENTERFORDATAINNOVATION

3

holdingsofmultiplepartiestoaddresscomplexquestionsandgeneratevaluableinsights.Thistypeofdata-sharingarrangementusuallyrequires

clearagreementstodefinedataaccessandusagerightsandthe

ownershipofintellectualproperty(IP),butspecificcharacteristicsmayvarydependingonthetypeofdatainvolvedandthenatureofthecollaboration.

Forexample,healthcareisonefieldinwhichpartnershipsbetween

organizationssuchashospitals,researchinstitutions,andmedical

providerscanhelpleveragedataanalyticsandartificialintelligence(AI)in

healthcareresearch,ultimatelyimprovingpatientoutcomesand

optimizingservicedelivery.

3

The23andMePatient-CentricResearchPortalallowscustomerstovoluntarilycontributetheirgeneticandself-reportedhealthinformationtoresearchstudies.

4

Thispartnershipbetween

23andMe,agenomicsandbiotechnologycompany,anditscustomers

enablestheadvancementofscientificunderstandingofvariousdiseasesandtraits.Researcherslinkgeneticdatainordertostudytopicssuchasancestry,traits,andevenrarediseases.

5

Data-sharingpartnershipshaveanumberofbenefitsforallpartners.Forresearchers,suchpartnershipsprovideaccesstogreaterdataforanalysisthantheywouldhaveontheirown,allowingforgreaterinsights.These

partnershipsalsohelpovercomethelimitationsofasingledatasetthatmaybetoosmallforcertaintypesofstatisticalanalysisorbemissing

relevantinformationneededforinvestigation.Infieldswheredataisoftenhighlysensitive,suchashealthcare,researchpartnershipsprotectthe

sensitivenatureofpatientinformationwhileallowinginstitutionstocollaborateandaggregateinsights.Suchpartnershipsalsomeanlessduplicationofdata,savingresearcherstimeandmoney.

Atthesametime,thisdata-sharingmodelhassomeconstraints.For

example,whendata-sharingpartnershipsarebetweentwocompetinginstitutions,thereareoftenIPandcompetitionconcerns.Likewise,suchcollaborationsmayinvolvedatasetsofvaryingqualityandstandards.

Theseissuesmustberesolvedbeforeanysharingcanoccur.

Recommendation:Facilitatedata-sharingpartnershipswithmodel

contracts.

Partnershipsbetweentwoentitiesarethemostbasicmodelofdata

sharingandshouldbesupportedbypolicymakers.Whenitcomestodata-sharingpartnerships,organizationsareoftenforcedtoreinventthewheelandgothroughanewcontractandnegotiationsprocesseachtimeanewopportunityforcollaborationcomesup.Policymakersinfederalagenciesshouldalleviatethisroadblockandfacilitatemoredata-sharing

partnershipsbydevelopingacontracttemplatethatorganizationscan

adoptandcustomize(e.g.,typeofdata,retentionterms,IPrights,etc.).

SomecountriessuchasSingaporealreadyprovidethistypeofguidancefordata-sharingpartnershipsandhaveacceleratedresearchandinnovationasaresult.

6

Moreover,theFederalTradeCommissionandDepartmentof

CENTERFORDATAINNOVATION

4

Justiceshouldprovideguidanceoncomplyingwithantitrustrulesoncollusionwhenusingthesemodelpartnershipagreements.

DataConsortia

Dataconsortiaalloworganizationstopooltheirdataforthebenefitofthegroup.

7

Whereasdata-sharingpartnershipsinvolvebilateralagreements,dataconsortiaconstituteaseriesofreciprocalsharingagreements.Theseconsortiacanexisttoaddressaspecificissueorforthegeneraland

ongoingexchangeofinformation.Forexample,agroupoftownsalongarivermightformadataconsortiumtosharedataaboutbacteriainthe

water,oragroupofhospitalsmightformadataconsortiumtosharedataaboutaspecificraredisease.Likewise,onlinemarketplacesmightformadataconsortiumtoexchangedataaboutthird-partysellersthataresellingcounterfeits.

8

Dataconsortiahavelongplayedaroleinfillingdatagaps.Forexample,theClinicalResearchDataSharingAllianceexiststoacceleratedrugdiscoverybysharingdatacollectedthroughouttheclinicaldevelopmentprocess.

9

Membersoftheconsortiumincludebiopharmaceuticalcompanies,

academicinstitutions,nonprofits,andpatientadvocacygroups,which

cometogetheraroundtheglobetoprovidecollectiveaccesstoclinicaldataandhelpdiversifystudypopulations.AnotherexampleoftheutilityofdataconsortiaistheLinguisticDataConsortium(LDC)attheUniversityof

Pennsylvania.

10

Thisgroupofuniversities,libraries,corporations,andgovernmentlabswasfoundedin1992“toaddressthecriticaldatashortagefacinglanguagetechnologyresearchanddevelopment.”

MembersoftheLDCsharelanguageresources,suchasspeechandtextdatabases,lexicons,andotherresources,thatplayabigroleintraininglargelanguagemodels.

Theprimarybenefitofdataconsortiaisthatitpromotesmoredatasharingandaggregation.Onlymembersofagivenconsortiumcanaccessthedata,andconsortiummemberstypicallymustcontributetothegroup.

Eventually,adataconsortiumwillcreateatipping-pointeffectinwhichitismorebeneficialtobeinthecollectivethanout.Onceatippingpointis

reached,aconsortiumensuresongoingdatasharingandgenerallypromotesapro-data-sharingworld.

Dataconsortiadohavesomedrawbacks.Beforeacriticalmassisreachedandatippingpointeffectoccurs,someorganizationsmightbebetteroffhoardingtheirdatafortheirexclusiveuse.Thismeansconsortianeedtoconsiderjoiningincentivesintheearlydaysofaneffort.

Recommendation:Surveyandidentifyopportunitiesforcross-sectordataconsortia.

Federalagenciesshouldcatalogdataconsortiathatexistwithinspecificsectorsandfacilitatethecreationofnewcross-sectorconsortiaforcriticalareas.Dataconsortiacanprovidepolicymakerswithaccesstoabroader

CENTERFORDATAINNOVATION

5

andmorediverserangeofdatasources,includingfromotheragenciesandtheprivatesector.Forexample,policymakersininterdisciplinaryagenciessuchastheFederalEmergencyManagementAgencyshouldcreate

consortiathatbringtogetherrelevantstakeholdersfromagenciessuchastheEnvironmentalProtectionAgency,theDepartmentofAgriculture,andtheDepartmentofHousingandUrbanDevelopmentaswellasprivate

sectororganizationstocoordinateongoingdatasharingandensuremorecoordinatedandeffectivedisasterresponse.

DataTrusts

Datatrustsareatypeofdatagovernanceframeworkthatmanage,protect,andsharedataforanagreedpurposeonbehalfofindividualsand

organizations.

11

Althoughtherecanbeconflictingdefinitionsofdatatrust,thecharacteristicsofthistypeofdata-sharingmechanismremainthe

same.Atthecoreofadatatrustisthedelegationofdatarightstoan

independentintermediary,knownasatrustee,whomakesdata-sharingdecisionswithresearchers,privatecompanies,andpublic-sectorbodiesthatbenefitthedatasubjects.

12

Datatrustsgivestructureandrulesformanagingandusingaggregateddataandhelpunlockitsvalueforthepublicinterest.

Datatrustsareanemergingmodelwithanumberofvariationsbeing

pilotedaroundtheworld.TheUnitedKingdomhastakenaparticular

interestinthedatatrustmodelforhealthcareapplications.Forexample,theUKBiobankmanagesthegenomicdataofmorethan500,000

individualswhohavedonatedtheirdataforuseinresearch.

13

Thedataisanonymizedandmadeavailabletoresearchersaroundtheworldto

acceleratescientificdiscoveryandimprovepublichealth.TheBiobankactsasatrusteeforthisdata—inotherwords,ithasafiduciaryresponsibilitytoholdandsharethedataforthebenefitsoftheUKpublic.Additionally,theUK’sNationalHealthService(NHS)isdevelopinganNHSFederatedDataPlatformtoaggregateallhealthdata,includingpersonalhealthrecords,

clinicaldata,andpublicdata,inonecentralizedplatformindividualsandtheprivatesectoralikecanaccess.

14

Thereareanumberofbenefitstodatatrusts,includingincreasingsocietalbenefitsfromdata,streamliningprocesses,andunlockingmorevaluefromdatabyenablingsecondaryuse.Overall,datatrustsareaninstitutionthatmultipleentitiescancontributetoandaccess,therebyfacilitatingongoingtransparencyandaccountabilityandconsistentrulesforthereuseofdata.Governmentscanthereforeaccessprivatesectordatainkeyareasunderaclearsetofagreements,andviceversa.InthecontextofAI,theycan

facilitateaccesstodiverseandhigh-qualitydatasets,enablingAI

developerstotrainandvalidatemodelsonmorecomprehensiveand

representativedata.Overall,datatrustsprovideatrustedframeworkformanagingdataresponsibly.

CENTERFORDATAINNOVATION

6

Atthesametime,datatrustsdocomewithcertainchallenges.Giventheoften-sensitivenatureofthedataheldbyatrust,datatrustscanbe

difficulttoimplementandsometimesaremetwithresistance.Alackof

socialtrustindatasharingcanleadtoprojectsbeinghelduporeven

canceled,suchasinthecaseofInBloom,aproposeddatatrustfor

educationthatwasmetwithsomuchstakeholderresistancethatitfailedtolaunch.

15

Moreover,datatrustscanberesource-heavy,requiringalotoffinancial,technical,andhumanresources.Lastly,datatrustscanbeat

oddswithdataprotectionlawsfocusedonsafeguardingindividualrights—whicharecommoninmanyWesterncountries—becausetheyfocuson

collectiveempowermentandbenefit.Thiscollectivemodelcanbedifficulttoimplementinthecontextofstringentdataprivacylaws.

Recommendation:Implementsector-specificdatatrusts.

TherearespecificdomainsintheUnitedStates,suchashealthcare,

transportation,education,andenvironmentalresearch,inwhichthe

establishmentofsector-specificdatatrustscouldprovidesignificant

benefitstosociety.Thesetrustscouldbringtogetherstakeholdersfrom

relevantsectorstopoolandgoverndata,enablingresearch,improving

servicedelivery,anddrivingsocietaloutcomesinspecificareas.

Consolidatingdataassetswithinaspecificsectorwouldenablea

comprehensiveunderstandingofsector-specificchallenges,trends,and

opportunities.Federalagencies,suchastheEnvironmentalProtection

Agency,theDepartmentofHealthandHumanServices,andthe

DepartmentofEducation,shouldestablishprogramstocreateandoperatedatatrustsintheirrespectivedomains.Byprovidingguidanceand

capacity-buildingsupport,federalagenciescouldhelpthedatatrustsnavigatethelegalandregulatoryframeworksspecifictoeachindustry.

DataCooperatives

Datacooperativesareaformofbottom-updatagovernanceinwhich

individualsvoluntarilypooltheirdatatonegotiatecollectivelywithprivatecompaniesandotherentities.Membersofadatacooperativeestablish

rulesondatasharingdesignedtobenefitthoseinthegroup.These

cooperativesoftenaimtomonetizemembers’collectivedataandare

fundedbytherevenuegeneratedfromdata-sharingagreements.Data

cooperativesaresimilartoagricultural,housing,andconsumercredit

cooperativesinwhichtheorganizationisownedandjointlymanagedbyitsmembers,whosharethebenefits.

Forexample,Driver’sSeatCooperativepoolsgigeconomyworkers’

smartphoneandmobilitydata,allowingthemtooptimizetheirincomes.

16

Thecooperativefunctionsthroughanappthatlinksmultiplesourcesofanindividualdriver’sdataandanalytics,thenaggregatesthatdataforall

membersofthecollective.Driver’sSeatalsosellsthesegroupinsightstolocalgovernmentslookingfordatatohelptransportationplanning

decisionsandsplitsdividendsamongmembers.Thistypeofdata-sharing

CENTERFORDATAINNOVATION

7

arrangementisprimarilydesignedtoempowerworkerstousetheirdataasacollectivebargainingmechanism.Datacooperativesalsoexistinthe

agriculturalsectorasameansofempoweringfarmerswithshared

knowledge.CooperativessuchastheNationalAgriculturalProducersDataCooperativeandtheGrower’sInformationServiceCooperativeoperateatthenationallevelandpooldatafromproducers,smallbusinesses,publicuniversities,andnonprofitsinordertoprovidefarmersandgrowerswithagriculturaldataandhelpenhancethesustainabilityoftheiroperations.

17

Onechallengeisthattheeconomicsofdatacooperativesdonotalwayswork.

18

Thevalueofeachindividualdatacontributormightberelativelysmall,butwithoutwidespreadparticipationfrommanydataholders,thecooperativewillfail.Datacooperativesthereforehavetocarefullychoosehowtocompensatemembers—toolittle,andnotenoughcontributorswilljoin;toomuch,anditisnotsustainable.

Datacooperativesarearelativenoveltyandmayhavelimitedapplications.However,justaslaborunionsprovideanimportantmechanismfor

collectivebargainingforworkers,datacooperativescanallowindividualstocollectivelynegotiatebenefitsfortheirdata.

Recommendation:Exploredatacooperativesinareaswherenondata

cooperativesandcollectivebargainingoccurs.

Datacooperativesareusefulwhenindividualsmaybereluctanttoshare

theirdatabecausetheyfearathirdpartywilluseitagainsttheirinterests,suchassmallfarmerswhoareconcernedthatlargeagricultural

companieswillusetheirdata,andinsightsfromtheirdata,toprofitattheirexpense.

19

Formingdatacooperativescangivethesedataholdersmore

negotiatingpowertosharedatawithothersontheirpreferredtermsandovercomereluctancetocollectandsharedata.Federalagenciesthat

alreadyprovideoversightorsupportforvarioustypesofnondata

cooperatives,suchastheU.S.DepartmentofAgriculture,theNational

CreditUnionAdministration,andtheNationalLaborRelationsBoard,

shouldconvenestakeholdersonthepotentialvalueofdatacooperativesintheirrespectiveareastopromotegreaterdatasharing.

FederatedDataAnalytics

Federateddataanalyticsisawaytoallowdataanalysistooccureven

whenorganizationsareunableorunwillingtosharetheirdata.Federateddataanalyticsreferstoadistributedapproachtodataprocessinginwhichdataisanalyzedindisparatelocationsandonlytheaggregatedinsightsarebroughttoacentralizedlocation.Forexample,acompanymightuse

federateddataanalyticstoanalyzedatastoredonitscustomers’deviceswithoutmovinganyofthatcustomerdatatothecompany.Instead,thecompanywouldonlyreceivetheresultsoftheanalytics.

20

Federateddataanalytics,includingmethodssuchasfederatedlearning,allowsdata

insightstobeobtainedwithoutsharingthedataitself.Bynotsharingthedata,thismethodcanassuagefearsaboutorganizationsaccessingor

CENTERFORDATAINNOVATION

8

storingsensitivedata,suchasconcernsaboutprivacyforindividualsorproprietarycompanyinformation.

21

Forexample,biopharmaceuticalcompanyBoehringerIngelheimand

precisionmedicinesoftwarecompanyLifebitBiotechhavepartneredto

buildascalablefederatedanalyticsplatformtocapturegenomicsinsightsfrombiobankdata.

22

Federatedanalyticsinthiscasepreservestheprivacy

ofthehighlysensitivebiomedicaldatabutstillallowsresearcherstoaccessinsightsfromindividualdatastoresandgeneratemedical

innovations.

Federatedanalyticshasanumberofbenefits,includingprovidinganewwaytoaccesslargequantitiesofdata.Forexample,biomedicalresearchandclinicaltrialsrequirepatientdatathatistypicallyheldbyanumberofdifferenthealthcareinstitutionsandboundbystrictprivacylaws.

23

Federatedanalyticsenablesprivacy-preservinganalysesofdatasets

withoutrevealinganyspecificpatientdata;andeachdataproviderretainscontrol.Thistypeofdatasharingcanenableprecisionmedicineandis

criticalinsituationswhereonedatasetfromoneproviderwon’tbeenoughtoidentifymeaningfulpatterns,suchasinthecaseofraredisease

research.

24

Eventually,federatedhealthdatanetworkscanfacilitatelarge-scaleanalysisacrossinstitutions,regions,andborders,apossibilitybeingconsideredbytheEU-U.S.TradeandTechnologyCouncil.

25

Atthesametime,federatedanalyticshasafewdrawbacks,including

problemsofcost,scalability,stakeholderresistance,andlackoftechnicalreadinessininstitutions.Asanemergingdata-sharingtechnique,

computingcostsforfederateddataanalyticsmaybehighandprohibitiveformanyapplications.Therearealsolimitedlarge-scaleexamplesof

federatedanalyticsatwork,particularlyinfieldssuchashealthcare,whichcreatesproblemsofscalabilitygiventhelackofablueprintforsome

sectors.Moreover,notallinstitutionsarereceptivetofederatedanalyticsortechnicallyequippedtoenableit.

26

Federatedanalyticsrequiresa

scalableplatformthatcandealwithlargequantitiesofdata,aswellasadvancedapplicationprogramminginterfaces(APIs)thatallowfor

coordinationbetweenplatforms.

27

Recommendation:ContinuefundingR&Dforfederatedanalytics.

TheWhiteHouseOfficeofScienceandTechnologyPolicyrecentlyreleaseda“NationalStrategytoAdvancePrivacy-PreservingDataSharingand

Analytics,”whichoutlinestheimportanceoffederatedanalyticsandotherprivacy-enhancingtechnologiestoimprovethevalueofdataforthepublicbenefitwhileprotectingindividualprivacy.

28

Whilesuchareportis

importanttoorganizingpolicyresponsetofederateddata-sharingmodels,federatedanalyticsshouldnotbeputonapedestalorconsideredasilverbullettodata-sharingdilemmas.Insomeways,federatedanalyticsexistsasatechnicalsolutiontoasocialproblemofdistrustinthecollectionanduseofdataorthelegalbarrierstodatasharingbetweendifferent

CENTERFORDATAINNOVATION

9

countries’dataprotectionlaws.However,federatedanalyticsdoesmakesenseasanoptionwhendatasharingisotherwisenotfeasible,so

policymakersshouldfundresearchanddevelopment(R&D)effortsinthisfieldtodevelopthiscapabilityanddeterminewhichsectorsstandto

benefitthemostfromfederatedanalytics.

CooperativeResearchandDevelopmentAgreements

CooperativeResearchandDevelopmentAgreements(CRADAs)are

partnershipsbetweenthegovernmentandprivatesectorinwhich

governmentresearchinstitutionssharedatawithprivatesectorpartnerstofacilitatethecommercializationofR&Dprojects.

29

Underthese

agreements,thegovernmentprovidespersonnel,services,facilities,IP,equipment,data,andmoretotheircollaborators—butnofunding.The

partners,whichcanbeprivatecorporations,nonprofits,universities,orevenstategovernments,providethesameservicesbutcanalsoprovidefundingfortheR&Defforts.

30

CRADAsallowprivatesectorpartnerstofileforpatentsandretainpatentrightswhileensuringthegovernmentgetsalicenseforanycommercializedproduct.

31

Forexample,theNationalOceanandAtmosphericAdministration(NOAA)usesCRADAstoleveragethevalueofitsdatasetsandbetterensurepublicaccesstodata.TheNOAABigDataInitiativebeganin2015toenlistprivatecompaniessuchasIBM,Microsoft,andAWStodevelopsolutionsto

increaseutilizationandaccesstoitsdata.

32

Duetobudgetaryandsecurityconstraints,NOAAcouldnotkeepupwithpublicdemandforaccessto

criticaldatasetsofthingssuchasweatherradardata,satelliteimagery,

historicalclimatedata,informationonfisheries,etc.

33

Theagencywantedtopromotetheuseofitsdataanddemocratizeaccess,whilecollaboratorshadtheinfrastructureexpertise.ThispartnershipreducedloadsonNOAAsystemsandbudgetsandcreatednewbusinessopportunitiesforthe

partnercompanies.

CRADAshaveanumberofbenefits.First,theyhavenoimpacton

governmentbudgets,astheyprimarilyutilizeexistinginfrastructureandsimplymakegovernmentandprivatesectorcollaborationmoreefficientandeffective.CRADAsalsoprotectIPandallowpartnerstomonetize

solutions,meaningtheyhavebuilt-inincentives.Atahighlevel,theseagreementshelpexpandandenhancetheexistingexpertiseof

government,industry,andacademiaandcontributetooverallnationalcompetitivenessandspurmoreinnovation.

CRADAsarehighlyfocusedandcontractualinnature,meaningtheyneedauthorizationandcarefulnegotiationoftermsrelatedtoliabilityandIP.

34

Forexample,federalresearchmustbemadeavailabletothepublic,so

CRADAsmustworkoutwaystowithholdresearchresultsforacertain

periodoftimeinordertoallowtheprivatepartnertopatentanyinventionsforcommercialuse.Whilefederalresearchinstitutionsareallentitledtoundertakethistypeofresearchagreement,notallmayhaveopportunities

CENTERFORDATAINNOVATION

10

forsuchcollaboration,orevenwantit.GiventhatCRADAsoccuronaproject-by-projectbasis,theydonotnecessarilyensureongoingpublic-privatepartnerships.

Recommendation:EvaluateareaswhereR&DcanbenefitfromCRADAarrangements.

ManygovernmentagenciesalreadyuseCRADAs,particularlyinthe

physicalandmedicalsciences,andfederalagenciesshouldcontinuetopromoteandauthorizethem.Giventhemanybenefitsofthistypeofdata-sharingagreement,CongressshouldasktheGovernmentAccountabilityOfficet