美國國土安全部：2024年人工智能路線圖

DEPARTMENTOFHOMELANDSECURITY

ARTIFICIALINTELLIG

ENCE

ROADMAP2024

Contents

LetterfromtheSecretary 4

LeveragingAItoAdvancetheDHSMission 5

AI-drivenChallengestoHomelandSecurity 6

ResponsibleUseofAI 7

ACoordinated,Government-WideApproach 8

WhoDoesWhatatDHS 9

DHSExternalEngagementonAI 11

OverviewoftheLinesofEffort

ResponsiblyLeverageAItoAdvanceHomelandSecurityMissions 13

PromoteNationwideAISafetyandSecurity 18

ContinuetoLeadinAIthroughStrong,CohesivePartnerships 22

Glossary 24

LetterfromtheSecretary

ArtificialIntelligencemaywellbethemostconsequentialtechnologyofourtime.Ithasthepowertoinnovatebeyondmeasure,andtoreshapehowwesecureournationandprotectourcommunities.AttheDepartmentofHomelandSecurity,weembracetheresponsibilityto

ensurethatAIisdevelopedandadoptedinawaythatrealizesitsfullpotentialwhileprotectingthepublicfromanyharmitsirresponsibleoradversarialusemightcause.

DHSisattheforefrontofemployingAIandmachinelearningtechnologiesandleadsbyexample.Withourtalentedanddedicatedteamof260,000personnelin22agenciesandofficesacrossthecountryandaroundtheworld,AmericansinteractdailywithDHSmorethanwithanyotherfederalentity.OuruseofAInotonlydeliversreal-timebenefitstothepublic,butalsofuelsourstrategiceffortsacrossallareasofhomelandsecurity.

WeareexcitedtointroducetheDHSAIRoadmap.ThisdocumentoutlinesourAIinitiativesandthetechnology’spotentialacrossthehomelandsecurityenterprise.ItisthemostdetailedAIplanputforwardbyafederalagencytodate,directingoureffortstofullyrealizeAI’spotentialtoprotecttheAmericanpeopleandourhomeland,whilesteadfastlyprotectingprivacy,civilrights,andcivilliberties.

ThisinitiativealignswithPresidentBiden’sExecutiveOrder14110,“Safe,Secure,and

TrustworthyDevelopmentandUseofArtificialIntelligence,”whichtasksDHSwithcriticalrolesto:

manageAIincriticalinfrastructureandcyberspace;

promotetheadoptionofAIsafetystandardsglobally;

reducethepotentialriskofAI’susetofacilitateweaponsofmassdestructionattacks;

combatAI-relatedintellectualpropertytheft;and,

ensureournationattractstalenttodevelopresponsibleAIintheUnitedStates.

OurroadmapforthecomingyearincludesexploringnewAIapplicationsandpursuingawhole-of-governmentstrategyforensuringthesafe,secure,andtrustworthydevelopmentanduseofAI.Weareseekingtoengagepartnersacrossthegovernment,privatesector,andacademiatobolsterournation’ssecurity.

Weinviteyoutocollaboratewithusinthisimportantmission.Together,wecanresponsiblyandethicallyleverageAItostrengthenournationalsecurity,improveouroperations,andprovideefficientservicestothepublicweserve.

AlejandroN.Mayorkas

AlejandroN.Mayorkas

U.S.DepartmentofHomelandSecuritySecretary

EricHysen

EricHysen

U.S.DepartmentofHomelandSecurityChiefInformationOfficer

4 DEPARTMENTOFHOMELANDSECURITY ARTIFICIALINTELLIGENCEROADMAP2024

LeveragingAItoAdvancetheDHSMission

CommittedtosafeguardingtheAmericanpeople,ourhomelandandourvalues,DHScontinuestoinnovateinsupportofitsmissions.

Whileitisnowfrequentlyinthenews,theconceptofAIhasbeenaroundsincethe1950s.Initiallyunderstoodasamachine’sabilitytoperformtasksthatwouldhavepreviouslyrequiredhumanintelligence,nowAIencompassessystemscapableofreasoning,inference,andlearning.ThesophisticationofAIsystemshasgrownsubstantiallyinthepastdecade,andparticularlyinthepastfewyears.AIsystemsareavailabletousersthroughinternet-basedinterfaces,increasinglyintegratedintosoftware,anddeployedbybusinessesandgovernmentsaroundtheworld.

DHShasusedAIforwelloveradecadeandcontinuestoincreasethebreadth,depth,andmaturityofAI’sapplicationacrosstheDepartment.Forexample,asearlyas2015,theDepartmentpilotedtheuseof

machinelearning(ML)technologiestosupportidentityverificationtasks.Sincethen,DHShassuccessfullyimplementedotherAI-poweredapplicationstoenhanceefficienciesandfosterinnovationinbordersecurity,cybersecurity,immigration,trade,transportationsafety,workforceproductivity,andotherdomainscriticaltoprotectingthehomeland.EveryDHScomponentandofficeisworkingtomeaningfullyassessthepotentialbenefitsofAItotheDHSmission,andtoresponsiblyharnessitspotentialtofurthertransformouroperations.TheDepartmenthasalreadypublished41differentusesofAIintheAIUseCaseInventoryat

/data/AI_inventory

,andthelistwillgrowasweexpanduseofAI.

Someexamplesinclude:

DHSisusingAItokeepfentanylandotherdangerousdrugsoutofourcountry.TheUnitedStatesCustomsandBorderProtection(CBP)usesanMLmodeltoidentifypotentiallysuspiciouspatternsinvehicle-crossinghistories.CBPrecentlyusedthemodeltoflagacarforsecondaryreview,whichyieldedthediscoveryofover75kgsofdrugshiddenintheautomobile.

DHSisusingAItoaidourlawenforcementofficersininvestigatingheinouscrimes.In2023,theUnitedStatesImmigrationandCustomsEnforcementHomelandSecurityInvestigationsOperationRenewedHopeidentifiedmorethan300previouslyunknownvictimsofsexualexploitationthanksinparttoanMLmodelthatenhancedolderimagestoprovideinvestigatorswithnewleads.

DHSisusingAIattheFederalEmergencyManagementAgency(FEMA)tomoreefficientlyassessdamagetohomes,buildings,andotherpropertiesafteradisaster.ThisapproachallowsFEMAinspectorstheabilitytolookatsomeimpactedstructuredamageremotelyinsteadofconductinginspectionsexclusivelyin-person,leadingtoswifterdeliveryofdisasterassistancetosurvivors.

DHSisusingAItomaketravelsaferandeasier.Byintroducingcustomer-facingtechnologiessuchasTouchlessCheck-Inattheairport,theTransportationSecurityAdministration(TSA)providestravelersanoptionalwaytonavigateTSAsecurityprocesses,checkbags,andboardtheirflightsbytakingjustaphotograph.Theseandothereffortsarealreadysavingtimeatsecuritycheckpointsandreducingphysicaltouchpoints.

TheseandothercurrentusesofAIareconductedinpartnershipandconsultationwiththeDepartment’s

OfficesofPrivacy,CivilRightsandCivilLiberties,GeneralCounsel,andotherappropriateoversightbodies.

DHSSecretaryAlejandroMayorkasparticipatesinafiresidechatatDEFCON2023.

AI-drivenChallengestoHomelandSecurity

WhiletherearetremendousopportunitiesforAItoenhancetheDHSmission,AIalsointroducesnewchallengesandrisks.TheproliferationofaccessibleAItoolslikelywillbolsterouradversaries’tactics.

CyberactorsuseAItodevelopnewtoolsthatallowthemtoaccessandcompromisemorevictimsandenablelargerscalecyber-attacksthatarefaster,moreefficient,andmoreevasive.Nation-statesseekingtounderminetrustinourgovernmentinstitutions,socialcohesion,anddemocraticprocessesareusingAItocreatemorebelievableforeignmaligninfluencecampaigns.

OfparticularconcernareimpactsofAIattacksoncriticalinfrastructures,whichcouldresultinnefariousactorsdisruptingordenyingactivitiesrelatedtoInternetofThings(IoT)technologiesornetworkedindustrialsystems.Generatingand

passingpoisoneddataintoacriticalsensorcouldtriggerdownstreamimpacts,suchasservicedisruptionsorsystemshut-offs.AIenabledtechnologiesarealsobeingusedtounderminethe

trustweplaceininformationderivedfromdigitalcontentandisdistinctfromtraditionalcybersecuritythreats,requiringadditionalresearchtounderstandandbuildknowledgetoinformprotections.Similarly,whileAIhasalreadyenabledinnovationinthephysicalandbiologicalsciences,italsohasthepotentialtosubstantiallylowerthebarrierofentryfornon-expertstodesign,synthesize,acquire,orusechemical,biological,radiological,ornuclearweapons.

CyberandphysicalsecurityisfoundationaltothesafetyandsecurityofAI.DHSandtheCybersecurityandInfrastructureSecurityAgencyinparticularwillcontinuetoworktoimprovethenation’soverallcyberresilienceandtoidentifyandmanagerisksassociatedwiththemisuseofAI/MLtechnologies.Additionally,assectorriskmanagementagencies,TSAandtheUnitedStatesCoastGuardwillcontinuetoassessAI-relatedrisksacrossthetransportationandmaritimesectors.

TSAusesnewcredentialauthenticationtechnologytoimprovecheckpointscreeningcapabilities.

ResponsibleUseofAI

DHSiscommittedtoensuringthatouruseofAI:

Isresponsibleandtrustworthy;

isrigorouslytestedtobeeffective;

safeguardsprivacy,civilrights,andcivilliberties;

avoidsinappropriatebiases;

andistransparentandexplainabletoour

workforceandtothoseweserve.

Whereappropriateitshouldbeinterpretabletomeetdueprocessrequirementsinlegaloradministrativeproceedings.DHS’suseofAIshouldadvanceequityandnotfunctioninwaysthatamplifyexistingsocialinequalities.

DHSpolicyalignswiththeDepartment’scommitmenttoleanforwardindeployingAItoolstoenhanceoperationsandleadthegovernmentintheresponsibleuseofAI.TheDepartment’s

governanceandoversightoftheresponsibleuseofAIarecloselycoordinated,highlycollaborativeeffortsthatuniteoperationalcomponentsandoversightofficesfromacrossDHSaroundthecommongoal

ofensuringresponsibleuse.Fromdayone,DHScomponentsandofficeshavecoordinatedcloselywithourPrivacyOffice,theOfficeforCivilRightsand

CivilLiberties,theOfficeoftheGeneralCounsel,andadditionalstakeholders.ThesecollaborationsensurecompliancewithallapplicablelawsandpoliciesandestablishclearguardrailstoprohibitinappropriateuseofAI/MLtechnologies,aswellassafeguardprivacy,civilrights,andcivilliberties.

In2023,DHSannouncednew

policiesand

measures

topromotetheresponsibleuseofAItoguidetheacquisitionanduseofAI/MLtechnologiesacrosstheDepartment.DHSappliestheprinciplesoftheblueprintforanAIBillofRightsandwillimplementtheNationalInstituteofStandardsandTechnology’sAIRiskManagementFrameworkwithintheDepartment.Asnewlawsandgovernment-widepoliciesaredevelopedandtherearenewadvancesinthefield,wewillcontinuetoupdateourinternalpoliciesandprocedures.

ACoordinated,Government-WideApproach

The

President’sExecutiveOrder14110

,whichestablishesacomprehensivestrategyfor

AIinnovation,directsDHStopartnerwithotherfederalagenciestodriveacoordinated,government-wideapproach.Specifically,DHSwillpartnerwithothersectorriskmanagement

agencies,suchassuchastheNationalInstituteofStandardsandTechnology,theDepartmentofJustice,andtheDepartmentofDefense,to:

Supportsector-specificAIriskidentification;

issueguidancetoagenciesonhowtolabelandauthenticatetheircontent;

performsecurityreviewsonAIfoundationalmodelsandothermeasures;

anddevelopguidelines,standards,andbestpracticesforAIsafetyandsecurity;

ThedirectionprovidedintheExecutiveOrderisconsistentwithDHS’sinnovativeworktoensurethesafe,secure,andresponsibledevelopmentanduseofAI.DHSwillcontinuetomanageAIincritical

infrastructureandcyberspace,promotetheadoptionofglobalAIsafetystandards,reducetheriskthatAIcanbeusedtocreateweaponsofmassdestructionandotherrelatedthreats,combatAI-relatedintellectualpropertytheft,andhelptheUnitedStatesattractandretainskilledtalent.

WhoDoesWhatatDHS

TheSecretarysetstheDepartment’sAIstrategy,priorities,andpolicies,andisthekeyinterlocutorwiththeprivatesector,federalinteragency,stateofficials,andkeyinternationalcounterparts.TheSecretaryrepresentstheDepartmentontheWhiteHouseAICouncil,whichcoordinatestheactivitiesofagenciesacrossthefederalgovernmenttoimplementAIpolicies.InApril2023,theSecretaryestablishedanAITaskForcewithinDHSthatdrivesspecificapplicationsofAItoadvancecriticalhomelandsecuritymissions.TheSecretaryhasalsoestablishedandwillchairanewAISafety

andSecurityBoard(AISSB),whichwillproviderecommendationsandadvicetotheSecretary,thecriticalinfrastructurecommunity,andthebroaderpubliconthedevelopmentanddeploymentofAI.IntheOfficeoftheSecretary,theSecretaryhasdesignatedtheSeniorCounselorforCybersecurityandEmergingTechnologytoserveastheExecutiveDirectoroftheAISSBandtoadvisetheSecretaryondeployingAIandotheradvancedtechnologiestofulfilltheDepartment’smissionandsupportrelatedpolicymakingacrosstheDepartmentandthroughouttheUSgovernment.

TheDepartment’sChiefInformationOfficer(CIO)servesasthedesignatedChiefArtificialIntelligenceOfficer(CAIO).TheCAIOpromotesAIinnovationandresponsibleuseacrosstheDepartmentanddevelops,inpartnershipwithDHSofficesandcomponents,theDepartment’sinternalpoliciesregardingitsuseofAI.TheCAIO

setsstrategicprioritiesforAIdeploymentsacrosstheDepartment,andonbehalfoftheSecretary,coordinatesAI-relatedeffortsinpartnershipswithDHSofficesandcomponents.TheCAIOalso

directstheDepartment’sinformationsecurity,datagovernance,informationtechnology,andcustomerexperiencefunctionswhichenableDHS’suseofAI.

TheScienceandTechnologyDirectorate(S&T),onbehalfoftheDepartment,isleadingAI-related

researchanddevelopmenttoprovidefederal,state,andlocalofficialswithcutting-edgetechnologyandcapabilitiestoprotectthehomeland.S&TleadsTest

&Evaluation(T&E)ofAI-enabledsystems,aswellasthedevelopmentofAIenabledT&EinpartnershipwiththePrivacyOfficeandoperationalcomponents.

TheOfficeofStrategy,Policy,andPlans(Policy)developstheDepartment’spolicyforAI,whichincludesoverseeingtheimplementationoftasksdirectedbyExecutiveOrder14110andenhancingcooperationwithinternationalalliesandpartnersonAIgovernanceandriskmitigation.PolicysupportstheSecretaryontheestablishmentandagendaoftheAISSB.

ThePrivacyOffice(PRIV)andtheOfficeforCivilRightsandCivilLiberties(CRCL)overseeandprovideguidanceontheresponsibleuseofAItoensuretheDepartment’suseofAIistransparent,explainable,trustworthy,avoidsinappropriatebiases,andsafeguardsprivacy,civilrights,andcivilliberties.PRIVoverseesDHS’suseofAItosafeguardpersonalprivacyandensurecompliancewithprivacypolicies.CRCLworkstopreserveindividualliberties,fairness,andequalitytoensurecompliancewithapplicableindividualrightsprotections,including(butnotlimitedto)dueprocessandnon-discriminationstandards,andworkstoadvanceequityacrossallDHSAIusecases.

TheCounteringWeaponsofMassDestruction(CWMD)Officeworkstopreventattacksusingweaponsofmassdestructionandotherrelatedchemical,biological,radiological,andnuclear(CBRN)threats,andleadstheDepartment’sworkinidentifyingandreducingrisksattheintersectionofAIandCBRNthreats.CWMDstrengthensfederalinteragencycoordinationandprovidesdirectfinancialaidandsupporttotheDepartment’soperationalcomponentsandstate,local,territorial,andtribalpartners,andfirstresponders.

TheCybersecurityandInfrastructureSecurityAgency(CISA)conductscyberdefensetoprotectagainstAI-enabledthreatsandsecureAI-basedsoftwaresystems,aswellasdriverelatedriskreductionandresilience.CISAisalsoresponsible

1In2023,CISAreleaseda

RoadmapforArtificialIntelligence

tooutlineassetofcomprehensiveactionsthatcompriseitsAImission.

HSIusesAItoinvestigateheinouscrimes.

forcommunicatingAI-relatedthreatandriskinformation,includingtocriticalinfrastructuresectors,andresponsiblyintegrateAIsoftwaresystemsacrossDHS.

OperationalComponents,includingtheTransportationSecurityAgency,UnitedStatesCitizenshipandImmigrationService,UnitedStatesImmigrationandCustomsEnforcementHomelandSecurityInvestigations,FederalEmergencyManagementAgency,andtheUnitedStatesCoastGuard,responsiblyintegrateAIintotheiroperationalcapabilitiestoharnessthepotentialofAI.Thesecapabilitiesincludeenhancedbordersecurity,counter-fentanylefforts,travelscreening,protectingintellectualproperty,grantandotherassistanceallocation,andmaritimesecurity.

TheOfficeofIntelligenceandAnalysisconductsandprovidesanalysisexamininghowAIimpactsthreats,includingthoserelatedtocybersecurity,foreignmaligninfluence,andcounterterrorism,tocriticalinfrastructureandtheHomelandSecuritymission.

ThePrivateSectorOffice,withintheOfficeof

PartnershipsandEngagement,fostersstrategic

communicationswithbusinesses,tradeassociations,andotherorganizationstocreatestrongerrelationshipswiththeDepartment.ThePrivateSectorOfficealsohelpsinformtheSecretaryontheimpactsofAIpoliciesandregulationsto

theprivatesector,aswellaspromotespublic-privatepartnershipsandbestpracticestodevelopinnovativeapproachestohomelandsecuritychallenges.TheOfficeofPartnershipsandEngagementwillprovideadministrativesupporttotheAISSBonbehalfoftheSecretary.

TheDHSAITaskForceincludesrepresentationfromacrossDHScomponentsandischairedbytheUndersecretaryforS&TandtheCIO.TheTask

ForcewasestablishedinApril2023todrivespecificapplicationsofAItoadvancecriticalhomelandsecuritymissions.IncollaborationwithCRCL,theTaskForceprovidesguidance,riskassessment,mitigationstrategies,andoversightfortheprotectionofindividualrights.TheTaskForcealsocoordinatesworktoaffectinternalDepartmentalpolicychangesandappliesoversighttoallDHSAIactivities.

DHSExternalEngagementonAI

DHSworksacrosssectorswithcriticalstakeholdersnationwideandconductsstrategicengagementsandoutreachwithstate,local,territorial,andtribalgovernments,electedofficials,theprivatesector,faith-basedandnon-governmentalorganizations,academia,andcommunitiesacrossthenation.TheDepartmentconvenesrepresentativesofthesestakeholdergroupsthroughvariouschannels,rangingfromstandingadvisorybodiestoworkshopsandothertargetedgatherings,toensuretheinterestsarerepresentedthroughDHS’spolicy-makingprocess.DHSalsosharesinformationanddevelopsrecommendationsorguidanceforincreasingsecurityandresilience.

SomeoftheDHS-convenedgroupsthatbringtogetheradiversegroupofstakeholderstoaddressAI-relatedissuesarelistedbelow:

TheAISSB,establishedbytheSecretaryatthedirectionofthePresident,pursuanttoExecutiveOrder14110ExecutiveOrderontheSafe,Secure,andTrustworthyDevelopmentandUseofArtificialIntelligence,tobringtogetherAIexpertsfromtheprivatesector,civilsociety,academia,andgovernment.ThepurposeoftheBoardistoadvisetheSecretary,thecriticalinfrastructurecommunity,andthebroaderpubliconthedevelopmentanddeploymentofAI;provideinformationandrecommendationsforimprovingsecurity,resilience,andsafety,includingpromulgatingspecificandactionableprinciples,guidelines,andbestpracticesfortheuseofAI;anddevelopeffectiveprocessestoreviewandrespondtoincidentsrelatedtotheuseofAIincriticalinfrastructure.

TheHomelandSecurityAdvisoryCommittee(HSAC)leveragestheexperience,expertise,andnationalandglobalconnectionsoftheHSACmembershiptoprovidetheSecretaryreal-time,real-world,andindependentadvicetosupportdecision-makingacrosshomelandsecurityoperations.In2023,theHSACissued

recommendations

onAItotheSecretary,whichhelpedinformmanyoftheactionsoutlinedinthisroadmap.

DHShasexistingstructuresinthecybersecuritydomainthatwillbecalledupontotheextentthatcybersecurityissuesariseinthecontextofAIuse.ChiefamongthemistheJointCyberDefenseCollaborative,whichbringstogethercyberdefendersfromorganizationsworldwidetogather,analyze,andshareactionablecyberriskinformationtoenablesynchronized,holisticcybersecurityplanning,cyberdefense,andresponse.IntheeventofasignificantcyberincidenteitherenabledbyAIoronAIsystems,theCyberSafetyReviewBoardwillreviewwhathappenedandproducespecificrecommendationstolimitthelikelihoodorimpactofsuchanincidentinthefuture.Additionally,

theCISACybersecurityAdvisoryCommitteeiscomprisedofexpertsoncybersecurity,technology,

riskmanagement,privacy,andresilience,whoadvisetheDirectorofCISAonpoliciesandprogramsrelatedtoCISA’smission,andspecifically,howtodriveadoptionofsecuresoftwaredesignpractices.

Otheradvisorypanelsandcommitteescanbefoundhere

DHSAdvisoryPanelsandCommittees

.

OverviewoftheLinesofEffort

Throughout2024,DHSwillfocusonthefollowinglinesofeffortandassociatedworkstreams:

ResponsiblyLeverageAItoAdvanceHomelandSecurityMissions

PromoteNationwideAISafetyandSecurity

ContinuetoLeadinAIthroughStrong,CohesivePartnerships

ResponsiblyLeverageAItoAdvanceHomelandSecurityMissionsWhileProtectingIndividuals’Privacy,CivilRights,andCivilLiberties

DHSwillleadacrossthefederalgovernmentintheresponsibleuseofAItosecurethehomelandanddefendagainstthemalicioususeofAI.WewillensurethatouruseofAIfullyrespectsprivacy,civilrights,andcivilliberties,isrigorouslytestedtoavoidprivacyharmsorimpermissiblebiasesandcanbeexplainedtothepeopleweserve.

WORKSTREAMONE

ContinuouslyandresponsiblypilotandimplementAItechnologiesinDHSmissionspaces

DHSwilluseAItosupportessentialfunctionsrequiredtosustainandsecuregovernmentoperations.GenerativeAItools,includingLargeLanguageModels(LLMs),couldbecomeintegralinreducingtheeffortexpendedonbusinessprocessessuchasmanaginghelpdesktickets.ThispragmaticapproachseekstoleverageAItostreamlineadministrativetasks,fosteringefficiencyandcost-effectivenessintheday-to-dayfunctioningofgovernment.AI,appliedstrategically,servesasatransformativetoolbyboostingproductivityandimprovingcustomerexperience.

Inadditiontothethreepilotslistedbelow,DHSexpectstoinitiateadditionalpilotsacrosstheenterprise,toincludeusinggenerativeAIforlanguagetranslation.

2024Goals

OfficerTrainingUsingLLMs

U.S.CitizenshipandImmigrationServices(USCIS)willpilotusingLLMstohelptrainRefugee,

Asylum,andInternationalOperationsOfficers

onhowtoconductinterviewswithapplicantsforlawfulimmigration.USCISwillgeneratedynamic,personalizedtrainingmaterialstosupplementhumantrainingthatadapttoofficers’specificneedstoensurethebestpossibleknowledgeandtrainingonawiderangeofcurrentpoliciesandlaws.Thepilotwillhelpenhanceofficers’

understandingandretentionofcrucialinformation,increasetheaccuracyoftheirdecisions,andlimittheneedforretraining.

PlanningAssistantforResilientCommunities

FederalEmergencyManagementAgency(FEMA)willpilotusinggenerativeAItoassistunderservedcommunitiesandlocalgovernmentswithdevelopinghazardmitigationplanstoidentifyhazards,assessrisksandvulnerabilities,anddevelopingmitigationstrategies.Approvedmitigationplansarearequirementforlocal

governments’eligibilityforprojectsfundedunderFEMA’sHazardMitigationAssistanceprogramandseveralothertypesofFEMAgrantsforbuildingcommunityresilience.Thepilotwillspecificallysupportstate,local,territorial,andtribalgovernments’understandingofhowtocraftaplanthatidentifiesrisksandmitigationstrategiesandhelpthosegovernmentsdraftplanelements—frompublicly-available,well-researchedsources—thattheycancustomizetomeettheirneeds.

EnhancedSearchandDocumentComprehensionusingLLMs

HomelandSecurityInvestigations(HSI)willpilotallowingofficerstouseLLMstosupporttheirinvestigativeprocessesby,(1)semanticallysearchmillionsofdocuments,(2)retrievingrelevant

casedocuments,and(3)providingofficersasummarizedresponsetoaspecificquerybasedonrelevantdocuments.Thesetoolsshouldenableinvestigatorstorapidlyuncoverkeyinformationandpatterns.TheseeffortsareinsupportofHSI’sworktocombatfentanyl,humantrafficking,childexploitationandothercriminalnetworks.

TransformBorderSecurity

U.S.CustomsandBorderProtection(CBP)willcontinuetoroll-outNon-IntrusiveInspectiontechnologytomakeborderscreeningsmoreefficientandtocombattherisksassociatedwithsmugglingfentanylandotherillicitgoods.Non-IntrusiveInspectionisacapabilityusedbyCBPtoconductsecondaryscreeningsatborderentrypointsviax-rayorimagingtechnologies,whichhelpsCBPidentifytheneedforanyadditionalmanualscreenings.CBPwillusedatagenerated

throughthesescreeningstofurtherenhancetheimagingtomoreaccuratelydetectanomalies.

EnhanceFederalCyberDefense

CybersecurityandInfrastructureSecurityAgency(CISA)willassesstheimpactofusingAI-enabledcapabilitiesforcybersecurityvulnerabilitydetectionandremediation.CISAwilldeliverareporttotheSecretaryofHomelandSecuritywithrecommendationsforactionsthatshouldbetakenbasedonlessonslearnedfromthisassessment.

WORKSTREAMTWO

BuildtechnicalinfrastructuretoacceleratesecureAIadoptionthroughoutDHS

2024Goals

AISandbox

OfficeofChiefInformationOfficer,incoordinationwithPrivacyOffice,OfficeforCivilRightsandCivilLiberties,andOfficeoftheGeneralCounsel,willlaunchanAISandboxforaninitialsetofDHSuserstoexperimentwithimplementingtheresponsibleuseofLLMsintheirsystems.TheaimistoexpandtheAISandboxtoadditionalDHSuserswithinoneyearandintegrateevolvingtestingandvalidationstandardsfittedtoDHSmissionandusecases.

InitialOperatingCapabilityforSURVEYORIntegratedDataEnvironment

UnitedStatesCoastGuard(USCG)willintroducetheSURVEYORIntegratedDataEnvironment,asuiteoftechnicalcomponentsservingspecificfunctionswithinanintegratedcloudarc