ResearchPaper

ACESintheHole?AutomatedCopyrightEnforcementSystemsandtheFutureofCopyrightLaw

ReginaZernay AdjunctFellow TheNiskanenCenter

RyanHagemann DirectorofTechnologyPolicy TheNiskanenCenter June6,2017

ExecutiveSummary Inthelastfewdecades,theartshaveseenanexplosionofcreativecontentthankstonewtoolsfordigitalcontentcreationanddistribution.Coupledwiththistremendousoutputofmaterialisacorresponding increase inbothpotentialandactualcopyright infringement.Yetdespitethesignificantlyincreasedriskofinfringingactivity,copyrightlawshavenotbeenmodifiedataratetomatch.As part of the effort to address this increased risk, the Digital Millennium Copyright Act(“DMCA”)was enacted. TheDMCA aims to serve the dual purpose of “promoting electroniccommerceandthedistributionofdigitalworksbyprovidingcopyrightownerswithlegaltoolstopreventwidespreadpiracy”whilestill“maintainingtheintegrityofthestatutorylimitationsontheexclusiverightsofcopyrightowners.”TheDMCAhasdonethisbyleavingthepolicingof

NiskanenCenter|2

online infringement up to the two groups most affected by online infringement—copyrightownersandtheonlineserviceprovidersthathostuser-generatedcontent.Each group’s ability to police infringement, however, has been drastically impacted by aninabilitytofiltertheenormousamountofmaterialconstantlyuploadedonline.Inresponsetothis dilemma, an interesting private-sector solution has emerged—the use of automatedsystems, or “content recognition systems” (“CRSs”), by private parties to identify and policeinfringement.Twoprimary types of automatedCRSs are currently in use. The first are automated systemsspecificallyusedforidentifyingpotentiallyinfringingworksandremedyingdisputationsthroughnon-legalchannels.Thesetypesofsystemsoffercopyrightownersseveralautomaticresponseoptions when infringement is detected: block the infringing content, permit it to remainavailableonthehost’swebsite,ormonetizeashareoftherevenue.Thesecondtypeofsystemsutilize automatedCRSs for the specific purpose of enforcing copyright protections. Examplesincludecontentrecognitionsoftwareusedbyuniversitiestoscanstudentpapersforplagiarism.To protect themselves, groups using content recognition software may set the software’sparameters at a very high level of responsiveness, sometimes going beyond the measuresrequiredbytheDMCA.TheeffectofthesemodificationsisthatCRSssometimesflagmaterialthatisnotactuallyinfringingunderthecurrentlegalstandard.Thus,thecurrentcopyrightenforcementsystemappearstobeacombinationofstatutesandprecedentmoresuitedtoapre-Internetworld,andaself-policingsystemthatmayimproperlyapply the law.Thispaperdiscusses the legal implicationsofmodifyingexistingprivate sectorCRSs to create a government-adopted automated copyright enforcement system that moreaccuratelyreflectsthelaw.Such an automated systemmayhelp resolve twoof thebiggest shortcomings in the currentcopyrightregime.First, itprovidesamoreappropriatesystemofevaluation,as itaddressesatechnology-basedproblemwitha technology-basedsolution.Second, it reshapes theexistingprivatesectorsolutionbycloselyapplyingthelegalstandardofinfringement,providingamoreaccurateevaluationofwhat is(andisnot) infringing. Inshort, itbridgesthegapbetweenthesystemsofcreationanddelivery,andthesystemofevaluation.Notwithstandingthepotentialbenefitsofsuchasystem,thispaperalsoacknowledgesthatthetechnicalcomplexitiesincreatingsuchasystemmaymakeitsimplementationdifficult.Italsorecognizes that, on net, the DMCA has been a benefit to both content creators andrightsholders, allowing for an explosive outgrowth of new content to dominate a newlyinterconnectedworld.Ultimately, this paper concludes that an automated copyright enforcement system mayeventuallybeaviablestepintheevolutionofcopyright law.However,thecreationofsuchasystem may rest in the distant future, largely due to the technical limitations inherent in

NiskanenCenter|3

accuratelydetectinginfringingcontent.Giventhecurrentecosystem,wearguethatpromotingCRSsandprivatesector,voluntarysolutions—allwithintheframeworkof theDMCA—isa farbetteralternativeforensuringthecontinuedprogressoftheusefularts.

NiskanenCenter|4

Introduction Overthelastfewdecades,technologyhasdevelopedatanastoundingrate.Newtechnologicalinnovationshaveenhancednearlyall fields, includingthose involvedwiththegenerationanddisseminationofcopyright-eligiblecontent.Theartshaveseenanexplosionoftoolsfordigitalcontent creation,digitizationofexistingworks, anddigitaldeliveryof content, encompassingeverythingfromthewrittenwordto3Dartandfilm.Copyright laws, however, have not beenmodified at a rate thatmatches the rapid influx ofdigitalcontent.1Whileexistingcopyright lawsmayhavebeenreasonablyeffectivebeforethetechnologicalrevolution,the law in itscurrentstatehasnotchangedquicklyenoughtomeettheincreasedriskofcopyrightinfringementinthedigitalworld.2Inanefforttoaddressthisissue,perhapsthemostsignificantadditiontocopyrightlawwastheDigital Millennium Copyright Act (“DMCA”).3 The DMCA has attempted to serve the dualpurposeof“promotingelectroniccommerceandthedistributionofdigitalworksbyprovidingcopyright ownerswith legal tools to preventwidespread piracy”while still “maintaining theintegrityofthestatutorylimitationsontheexclusiverightsofcopyrightowners.”4OnewaytheDMCA achieves these purposes is by offering online content hosts, or “online serviceproviders”5 (“OSPs”), safe harbor protections from secondary liability for copyrightinfringementwhentheirusersuploadinfringingcontent.Toreceivethisprotection,OSPsmustfollowa set of specific steps, including the “expeditious” removal of infringing content uponnotification.6Removing infringingcontentsoonafterdiscovering itsexistence isnotassimpleas itsounds.Thedifficultystems fromthesheervolumeofuser-generatedmaterialhostedbyonlinesiteslike YouTube, Facebook, Vimeo, and others. For example, in July 2015, YouTube CEO Susan

1See,e.g.,NavinKatyal,TheUnauthorizedDisseminationofCelebrityImagesontheInternet...intheFlesh,2TUL.J.TECH.&INTELL.PROP.1(2000)(“[T]heInternetposesanewthreatforcopyrightholders,becausetechnologyseemstohaveoutpacedcurrentcopyrightlaws.”).2See,e.g.,VitoA.Costanzo,BuildinganIPLitigationandContentProtectionStrategyforClientswithInternet-RelatedCopyrightandTrademarkIssues,inMANAGINGINTELLECTUALPROPERTYISSUESINCYBERSPACE:LEADINGLAWYERSONDEVELOPINGANEFFECTIVEINTERNETIPSTRATEGY(2012)(“TheInternetnotonlymakesitmucheasiertoaccess,duplicate,anddistributecopyrightedmaterial,italsomakesiteasiertosueforcopyrightinfringement.”).3ExecutiveSummary:DigitalMillenniumCopyrightAct,Section104Report,U.S.COPYRIGHTOFFICE,https://www.copyright.gov/reports/studies/dmca/dmca_executive.html(lastvisitedJan.14,2017)(“TheDigitalMillenniumCopyrightActof1998(DMCA)wasthefoundationofaneffortbyCongresstoimplementUnitedStatestreatyobligationsandtomovethenation’scopyrightlawintothedigitalage.”).4Id.5TheDMCAdefinesa“serviceprovider”as“anentityofferingthetransmission,routing,orprovidingofconnectionsfordigitalonlinecommunications,betweenoramongpointsspecifiedbyauser,ofmaterialoftheuser’schoosing,withoutmodificationtothecontentofthematerialassentorreceived.”17U.S.C.512(k)(2011),https://www.copyright.gov/title17/92chap5.html#512.617U.S.C.512(c)(2011),https://www.copyright.gov/title17/92chap5.html#512.

NiskanenCenter|5

Wojcicki“revealedthat400hoursofvideoarenowuploadedtoYouTubeeveryminute.”7Tobesure, this remarkable wealth of content should be viewed as a valuable benefit to society,demonstratingthatthetechnologicalrevolutionhasledtotheprogressofusefularts.Despitenewfounddifficultiesincopyrightmanagementandenforcement,it isundeniablethatweareliving in an age of unparalleled content creation. The digital age has seen a blossoming ofcontent online—from open source software development and self-published authors tohomebrewedvideoproducersandindependentmusiciansreachingwideraudiencesthantheyeverwouldhavebeenabletobeforetheadventoftheInternet.Farfromstymyingcreativity,thecurrentsystemhasallowedcontenttobloominawayneverbeforethoughtpossible.MuchofthathasbeenduetothecompromisesdrivenbytheDMCA.However,thevastamountofuser-generatedmaterialuploadedonlinehasmadefilteringoutinfringing contentmore challenging for the parties affected. Two groups propel the DMCA’sprocedure for identifyingand removing infringingonline content—copyrightowners (or theirrepresentatives)andtheOSPsthemselves.8Eachgroup’sabilitytopoliceinfringingcontenthasbeenaffectedbytheenormousamountofmaterialhostedonline.9Thoseenforcingtherightsofcopyrightownersmaybe incapableofmanuallysearching foreveryOSPhosting infringingcontent.10 By the same token, if OSPs experience a dramatic increase in notifications aboutinfringing content, theymay not have the resources tomanually evaluate and expeditiouslyremovethecontent.11Interestingly, a private-sector solution has emerged—the use of automated systems.12 Thispaperwillrefertothosesystemsmoregenerallyascontentrecognitionsystems(“CRSs”).Forthesakeofclarity,thispaperwillrefertothesoftwareincorporatedintothesesystemssimplyas “content recognition software.” One example includes the popular smartphone appShazam.13Shazamusessmartphones to“listen” toa fewsecondsofa songand identify theartist and song title.14 CRSs are capable of identifying, with various degrees of technicalaccuracy, specific content. This technology, applied in an automated fashion to the realmofcopyright,isthetopicofthispaper.

7GregJarboe,VidCon2015Haul:Trends,StrategicInsights,CriticalData,andTacticalAdvice,TUBULARINSIGHTS(July27,2015),http://tubularinsights.com/vidcon-2015-strategic-insights-tactical-advice/.8JENNIFERM.URBAN,JOEKARAGANIS&BRIANNAL.SCHOFIELD,NOTICEANDTAKEDOWNINEVERYDAYPRACTICE31(2016),https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2755628.9Id.10Id.11Id.at31–32.12Id.;MichaelW.Carroll,PinterestandCopyright’sSafeHarborsforInternetProviders,68U.MIAMIL.REV.421(2014)(“Inrecentyears,somecopyrightownershaveinvestedinautomatedenforcementbyrelyingoncomputeralgorithmstoidentifythepresenceofacopyrightedwork...andtosendanautomatedtake-downnoticetoserviceproviders....Someserviceproviders,suchasYouTube,haveinvestedinautomatedprocessestorespondtosuchnotices.”).13Shazam–NameAnySonginSeconds,SHAZAM,https://www.shazam.com/(lastvisitedMar.21,2017).14AboutUs,SHAZAM,https://www.shazam.com/company(lastvisitedMar.21,2017).

NiskanenCenter|6

TherearetwotypesofautomatedCRSsdiscussedinthispaper.Inordertoavoidconflatingthetwo, theiroperationsandeffectsneed tobeclearlydifferentiated.Both typesutilizecontentrecognitionsoftwareasidentificationmechanisms,butaimtoachievedifferentobjectives.Thefirstareautomatedsystemsspecificallyusedforidentifyingpotentiallyinfringingworksandremedying disputations through non-legal channels, which we will refer to as automatedidentification and remediation systems (“AIRSs”). Once potentially infringing content isidentified, these systems invokeapre-chosenprocess selectedby the rightsholder:block theinfringingcontent,permitittoremainavailableonthehost’swebsite,ormonetizeashareoftherevenue(oftenthroughassociatedadvertisements,asinthecaseofYouTube’sContentIDsoftware). These systems have been developed by private sector firms operating under theconstraints imposedby theDMCA.Theirvoluntaryadoptionhas largelyhelpedeasemanyofthecontentious issues thatwouldotherwisehaveexistedbetweenrightsholdersandcontentcreatorsinthedigitalage.AIRSs are already in useonline for a variety of purposes. For instance, every time a video isuploadedtoYouTube,thevideoisscannedforthepresenceofcopyrightedmaterialownedbyotherusers.Ifpotentiallyinfringingmaterialisfound,thecontentmaybemonetized,blocked,orleftup,asdeterminedbytheowneroftheinfringedcontent.15The second type utilizes automated CRSs for the specific purpose of enforcing copyrightprotections. Examples of the second type of automated system include content recognitionsoftwareusedbyschoolsanduniversitiestoscanstudentpapersautomaticallyforplagiarism.16Thispaperwill refer to theseas automated copyright enforcement systems (“ACESs”). ThesearedistinctfromAIRSsinthattheultimategoalofanACESisto“enforce”copyrightclaims,notremedydisputesthroughvoluntaryornon-legalchannels.ThoughCRSsareusedformanypurposes,veryfewdetailedexplanationsofhowtheyworkareavailable.17Perhapsthemostcomprehensivedescriptionisfoundinarecentreportreleasedbythe non-profit group Engine, entitled The Limits of Filtering: A Look at the Functionality &Shortcomings of Content Detection Tools (“The Limits of Filtering”).18 The Limits of Filteringdescribes threemethodsusedtodeterminewhethercontent is infringing:metadata filtering,hash-basedfiltering,andcontentfingerprinting.19AsTheLimitsofFilteringnotes,infringerscan

15HowContentIDWorks,YOUTUBE,https://support.google.com/youtube/answer/2797370?hl=en(lastvisitedMar.21,2017).16ProfessorsUseTechnologytoFightStudentCheating,U.S.NEWS&WORLDREPORT(Oct.3,2008,5:26PM),https://www.usnews.com/education/articles/2008/10/03/professors-use-technology-to-fight-student-cheating.17URBAN,KARAGANIS&SCHOFIELD,supranote8at3(“Somesecrecyisdefensible.Rightsholdersexpressedconcernthatpublicizingdetailsoftheirenforcementpracticesmightenablepiratestocircumventthem.OSPsworriedthatrevealingtheirpracticescouldsubjectthemtonegativeattentionfromrightsholders,targets,orotherOSPs,forcingthemtochangetheirpractices.”).18SeeEVANENGSTROM&NICKFEAMSTER,THELIMITSOFFILTERING:ALOOKATTHEFUNCTIONALITY&SHORTCOMINGSOFCONTENTDETECTIONTOOLS(2017),http://www.engine.is/the-limits-of-filtering.19Id.ati–ii.

NiskanenCenter|7

avoid detection under the first method, metadata filtering systems, by simply altering themetadata informationofthe infringingfile.20Similarly,evenminoralterationsofan infringingfile,suchasshorteningthefileorconverting it toadifferent filetype,maycausethesecondmethod, hash-based filtering systems, to miss a match.21 Since the third method, contentfingerprinting, more closely evaluates the actual content of files,22 this paper’s analysis willfocusonthatmethod.One caveat: It is significant to note that cryptographic hash functions are often used togeneratedigitalfingerprintsoffiles,andasaresult,somecontentfingerprintingsystemscouldbeconsideredhash-based.23Thoughcryptographichashfunctionsareatypeofhashfunction,they are described as a “more powerful” formof hash-based filtering, and differ fromotherhash functionsbecause theyaremuch less likely to result in twodifferent files returning thesame output.24 The Limits of Filtering makes a further distinction, remarking that becausecontentfingerprintingsystems“relyonalgorithmsthatprocesstheunderlyingmediacontentof a given file, they are naturally constrained to a small subset of copyrightable content.”25Thus, while some refer to cryptographic hash functions as simply “hash functions,” this istechnically incorrect, since the term “hash functions” serves as a generic description thatencompassesbothcryptographichashfunctions(likethoseusedforcontentfingerprinting)and“othersortsofalgorithmslikecyclicredundancychecks.”26Whilesomedescriptionsofcontentfingerprintingmayconfusinglyrefertotheuseof“hashes,”thislikelyreferencestheparticularsubsetofcryptographichashfunctions,andnothash-basedsystemsingeneral.Onabasiclevel,mostcontentfingerprintingsoftwaresystemsperformthefivefollowingsteps:

(1) Copyrightedmaterialissubmittedforinclusioninthesoftware’sdatabase;(2) The software analyzes the material for distinctive features and creates a “digital

fingerprint”ofthematerialbasedonthisanalysis;(3) Thedigitallyfingerprintedmaterialisaddedtothesoftware’sdatabase;(4) Newmaterial being evaluated for amatch is similarly scanned and compared against

materialinthesoftware’sdatabase;

20Id.at12.21Id.at13.22Id.at13–14.23See,e.g.,SimsonGarfinkel,FingerprintingYourFiles,MITTECHNOLOGYREVIEW(Aug.4,2004),https://www.technologyreview.com/s/402961/fingerprinting-your-files/;DavidAshfield,WhyAreCryptographicHashFunctionsImportantinDigitalForensics?,CCLGROUP(Jan.30,2014),https://www.cclgroupltd.com/cryptographic-hash-functions-important-digital-forensics/;WhatisCryptographicHashing?MD5,SHA,andMore,TIPTOPSECURITY(Dec.15,2014),https://tiptopsecurity.com/what-is-cryptographic-hashing-md5-sha-and-more/.24JOHNEDWARDSILVA,ANOVERVIEWOFCRYPTOGRAPHICHASHFUNCTIONSANDTHEIRUSES3(2003),https://www.sans.org/reading-room/whitepapers/vpns/overview-cryptographic-hash-functions-879.25ENGSTROM&FEAMSTER,supranote18.26TimFisher,CryptographicHashFunction,LIFEWIRE(updatedOct.30,2015),https://www.lifewire.com/cryptographic-hash-function-2625832.

NiskanenCenter|8

(5) Ifthesoftwarerecognizesenoughofasimilaritybetweenthenewmaterialandmaterialinitsdatabase,amatchisidentifiedandflagged.27

Whatconstitutesa“match”dependsonhowthesoftwarehasbeenprogrammed.Factorssuchasthetypesandpercentageofsimilaritiesserveasparameters forthesoftware’sevaluation.Theparametersusedandtheirlevelofsensitivityarefeaturesusuallyinfluencedbytheneedsoftheuserlicensingorcommissioningthesoftware.28Toprotectthemselves,groupsusingcontentrecognitionsoftwaremaysetparametersataveryhigh level of responsiveness. For example, some OSPs have adopted automated proceduresthat “go beyondmeasures required by section 512 [of the DMCA].”29 This type of cautiousbehaviorhasbeenobservedbybothrightsholdersandOSPsattemptingtopreservetheirrightsandprotectionsundertheDMCA.Theeffectofthesemodificationsisthatsystemssometimesflagmaterialthatisnotactuallyinfringingunderthecurrentlegalstandard.Atleastonereportnotedseveralissuesresultingfromthisincorrectapplicationofthelaw,finding“nearly30%of[automated] takedown requests were of questionable validity.”30 This suggests thatrightsholdersandOSPsareengaging inasystemofdefactoself-policingthat is lesseffectivethanitcouldbe,givenprevailinglegalstandards.31HarmscausedbyerroneousDMCAtakedownnoticesresultingfromthisself-policinginclude:

• Clearlymismatchedmaterial. For instance,anoticewassentonbehalfofParamountaccusing Paramount’s Anchorman: The Legend of Ron Burgundy of infringing theParamountmovieAnOfficerandaGentleman.32

• Failure to consider fair use or “legitimate political speech.” NBC issued a takedownnotice for a “satirical Obama campaign video, which used archival footage of TomBrokaw to announce the ‘bad news’ that Senator John McCain had been electedpresident.” The videowas “[i]ntended tomotivateObama supporters to ‘get out thevote,’”but“wastakendowndaysbeforeacriticalvoterregistrationdeadline.”33

• Improper takedowns caused by mistaken claims of ownership. In 2012, YouTubefootage of NASA’s Curiosity rover landing onMars, a public domain video posted byNASAon its ownYouTubepage,was removedbecauseof a takedownnotice sent byScrippsNewsServiceincorrectlyclaimingScripps’ownershipofthevideo.34

27JonathanStrickland,HowContent-RecognitionSoftwareWorks,HOWSTUFFWORKS(July16,2007),http://computer.howstuffworks.com/content-recognition.htm.28Id.29URBAN,KARAGANIS&SCHOFIELD,supranote8at1–2.30Id.at3.31Id.at2–3.32Id.at91.33NBCIssuesTakedownonViralObamaAd,ELECTRONICFRONTIERFOUNDATION,https://www.eff.org/takedowns/nbc-issues-takedown-viral-obama-ad(lastvisitedMar.20,2017).34AlexanderHiggins,NASA’sMarsRoverCrashedintoaDMCATakedown,BEFOREIT’SNEWS(Aug.7,2012,1:48PM),http://beforeitsnews.com/alternative/2012/08/nasas-mars-rover-crashed-into-a-dmca-takedown-2448814.html.

NiskanenCenter|9

• Abuseof theDMCANoticeandTakedownprocess. Inonecase, apolitical candidateallegedlyusedtheDMCAtoshutdowntheYouTubeaccountofPeoplefortheAmericanWay’s RightWingWatch (“RWW”) by sending “a series of takedownnotices claimingthatRWW'suseofclipsofhisprograminfringedhiscopyrights,”though“RWWwas[]postingthemaspartofitscriticalwork.”The“boguscomplaints”causedRWW’sentireYouTubeaccount“tobetakenoffline—twice.”35

Theseandotherexamplessuggestthecurrentcopyrightenforcementsystemisacombinationofstatutesandprecedentmoresuitedtoapre-Internetworld,andaself-policingsystemthatoftenimproperlyappliesthelaw.Thispaperdiscusses the legal implicationsofagovernment-adoptedACES.Basedonsystemsthat already exist, but modified for the purpose of more closely applying the law, such anautomatedsystemmayhelpresolvetwoofthebiggestshortcomings inthecurrentcopyrightsystem.First,itprovidesamoreappropriatesystemofevaluation,asitaddressesatechnology-basedproblemwithatechnology-basedsolution.Second,itreshapestheexistingprivatesectorsolution by closely applying the legal parameters of infringement, offering a more accurateevaluationofwhatis(andisnot)infringing.Inshort,itbridgesthegapbetweenthesystemsofcreationanddelivery,andthesystemofevaluation.However,itisworthnotingthatthetechnicalcomplexitiesincreatingsuchasystemmaymakeitsimplementationdifficult.WealsoarguethattheDMCAhas,onnet,beenabenefittobothcontent creators and rightsholders, allowing for an explosive outgrowth of new content todominateanewlyinterconnectedworld.Asaresult,wenotethatthetechnologytomaintainarepositoryofvariouscopyrightedcontentmay,inthefuture,helpbetteridentifyrightsholdersand assist in detection of infringing materials. In the meantime, however, the existing safeharborsystem,buttressedbyinnovativetechnologicalsolutionslikeCRSs,isworkingwell,andany changes to the system should aim tomaximize the options for rightsholderswhen theirworksareusedbyothers.Part I of this paper describes the current state of the copyright enforcement system. Part IIexamines CRSs currently in use by the private sector. Part III considers potential uses for anACES.Part IVoutlinesrecommendedstepstodesignandimplementanACESforgeneraluse.Finally,PartVaddressespotentialobstaclestoimplementation.Ultimately,thispaperconcludesthatanACESmayeventuallybeaviablestepintheevolutionof copyright law. The structure of an ACES has already been created through private sectordevelopment. Thus, its implementation couldpossiblybe achieved through the licensing andmodificationofexistingautomatedsystems.However,thecreationofsuchasystemmayrestinthe distant future, largely due to the technical limitations inherent in accurately detectinginfringing content. Given the current ecosystem, we argue that promoting CRSs and private35AttempttoSilencethePoliticalSpeechatRightWingWatch,ELECTRONICFRONTIERFOUNDATION,https://www.eff.org/takedowns/attempt-silence-political-speech-right-wing-watch(lastvisitedMar.20,2017).

NiskanenCenter|10

sector,voluntarysolutions—allwithintheframeworkoftheDMCA—isafarbetteralternativeforensuringthecontinuedprogressoftheusefularts.

I.TheCurrentStateofCopyrightEnforcement The current copyright enforcement system leaves a great deal of uncertainty about whatcontent copyright owners actually possess and howwell this contentwill be protected. Thisuncertaintyenhancestheattractivenessofanautomatedsystemthatconsistentlyappliesthelaw,whichmayleadtomorepredictableoutcomes.Between the long duration of copyright ownership36 and the unclear line separatingcopyrightablecontentfromthepublicdomain,37itisoftenunclearwhatmaterialisprotectable.Additionally, thestandardfor infringementhasbeen inconsistentlyapplied,making itdifficultforcopyrightownerstodeterminewhatconstitutesaviolation.38Forthosewhoowncopyrights,itisdifficulttoprotectandfullyexploittheirproperty.Forthosewhohavecreatedprofitablenewcontent,earningsarejeopardizedbythehigherlikelihoodofinfringementsuits.Those issues,however,mustnecessarilybecomparedto theconsiderablebenefits associated with the recent content explosion made possible by the era of digitalcommunications.IntheyearsfollowingthepassageoftheDMCA,theglobaleconomicvalueof

36Currently,thedurationofthecopyrighttermforsingleauthorscoversthelifeoftheauthorplus70years.17U.S.C.302(2011).Foramorein-depthdiscussionofhowcopyrightdurationhasincreasedovertime,seeChapter6ofTOMW.BELL,INTELLECTUALPRIVILEGE:COPYRIGHT,COMMONLAW,ANDTHECOMMONGOOD(2014).SeealsoPaulBelleflamme,TheEconomicsofCopyrightProtection,IPDIGIT(Oct.2,2013),http://www.ipdigit.eu/2013/10/the-economics-of-copyright-protection/.37Forinstance,Copyright&FairUsewarnsofseveraltypesofworksthat,whilebasedonpublicdomainmaterials,arenotinthepublicdomainduetoanynumberofreasons,includingmodification,protectionundertrademarklaw,orprotectioninothercountries.SeeRichStim,PublicDomainTroubleSpots,COPYRIGHT&FAIRUSE:STANFORDUNIVERSITYLIBRARIES,http://fairuse.stanford.edu/overview/public-domain/trouble-spots/(lastvisitedJan.6,2017).38See,e.g.,RachelIsabelleButt,AppropriationArtandFairUse,25OHIOST.J.ONDISP.RESOL.1055(2010),https://kb.osu.edu/dspace/bitstream/handle/1811/76940/OSJDR_V25N4_1055.pdf?sequence=1(“Thecurrentsystemtodetermineafairusethroughlitigationisineffectiveasappliedtoconflictsinvolvingvisualartsbecausethecourtshavemisusedthefairusedoctrine.Thefirstproblemisthatinconsistentcaselawregardingfairuseandcopyrightinfringementfailstoprovideguidanceforartists.Thesecondproblemisthatthejudicialconceptofcopyrightinfringementdoesnotcomportwiththeacceptednormsintheartworld;artistshistoricallyhaveborrowedandcopiedexistingexpressionwithoutobjectionorconflict.”);TonyaM.Evans,Sampling,Looping,andMashing...OhMy!:HowHipHopMusicisScratchingMoreThantheSurfaceofCopyrightLaw,21FORDHAMINTELL.PROP.MEDIA&ENT.L.J.843,847–48(2011),http://ir.lawnet.fordham.edu/iplj/vol21/iss4/1(“differentinfringementstandardsarebeingappliedbythecircuitcourtstosoundrecordinginfringementcases....[t]heresultingincongruentdecisionsreflectaninconsistentapplicationoffederallaw.”);MarkA.Lemley,OurBizarreSystemforProvingCopyrightInfringement,57J.COPYRIGHTSOC’YU.S.A.719,741–42(2010)(criticizingtheexistingstandardfordemonstratingcopyrightinfringementin2010,stating“[o]urrulesforprovingcopyrightinfringementmakelittlesense.”);PamelaSamuelson,AFreshLookatTestsforNonliteralCopyrightInfringement,107NW.L.REV.1821,1823(2013),http://scholarlycommons.law.northwestern.edu/cgi/viewcontent.cgi?article=1045&context=nulr(“Onereasonwhyconventionaltestsforjudgingnonliteralcopyrightinfringementareproblematicisthattherearetoomanytestsandnotenoughguidanceaboutwhichonetouseinwhatkindsofcases.”).

NiskanenCenter|11

the entertainment industry has grown by 66 percent, “andU.S. households spent almost 15percentmoreonentertainmentasapercentageofincomein2008thanin2000.”39AsnotedinThe Limits of Filtering, “[i]n terms of actual content creation, the years since the DMCA’spassagehavebeenamongstthemostprolificinhistory.[OSPs]haveseensimilargrowth,asthe[I]nternet sector contributednearly$1 trillion to theU.S. economy—nearly6percentof realGDP—in2014.”40Nonetheless,legaluncertaintypersistsinthecopyrightrealm.Adiscussionofthecurrentstateofcopyrightenforcementisthereforeworthdelvinginto.What follows is a discussionof factors that have contributed to theuncertainty of copyrightownership, concerns regarding regulation of online distribution, and the current self-policingsystem.

A. TheUncertaintyofCopyrightOwnershipThere are a number of factors that contribute to the uncertainty of rights possessed bycopyrightowners.Itisnotalwaysclearwhatisprotectedbycopyright.Thecourtshaveacknowledgedthat“[t]hedeterminationof[copyright] infringement isoneofthemostdifficultofall legalquestions,”41andatleastoneobserverhasnotedthatthecourts’rulesforevaluatingcopyrightinfringement“makelittlesense.”42It is also unclearwhat is copyrighted andwhatmay be freely used because it is part of thepublicdomain.Materialinthe“publicdomain”hasbeendefinedas“creativematerialsthatarenotprotectedby intellectualproperty lawssuchascopyright, trademark,orpatent laws.Thepublicowns theseworks,notan individual authororartist.Anyonecanuseapublicdomainwork without obtaining permission, but no one can ever own it.”43 However, determiningwhethermaterial fallswithin the public domain is challenging, since the answer depends onwhentheworkwascreated,whetherandwhentheworkwaspublished,andwhethertheworkinquestionincludesmaterialfromanearlierwork.44

39ENGSTROM&FEAMSTER,supranote18,at9;seealsoJoshuaP.Friedlander,NewsandNoteson2015RIAAShipmentandRevenueStatistics,https://www.riaa.com/wp-content/uploads/2016/03/RIAA-2015-Year-End-shipments-memo.pdf(lastvisitedApril9,2017).40ENGSTROM&FEAMSTER,supranote18,at9–10.41MichelleV.Francis,MusicalCopyrightInfringement:TheReplacementofArnsteinv.Porter–AMoreComprehensiveUseofExpertTestimonyandtheImplementationofan“ActualAudience”Test,17PEPP.L.REV.493,494(1990)(citingA.SHAFTER,MUSICALCOPYRIGHT146(2ded.1932)).42Lemley,supranote38,at741–42.43RichStim,WelcometothePublicDomain,STANFORDUNIVERSITYLIBRARIES,http://fairuse.stanford.edu/overview/public-domain/welcome/(lastvisitedJan.14,2017).44SeeOfficeoftheGeneralCounsel,CopyrightandFairUse,HARVARDUNIVERSITY,http://ogc.harvard.edu/pages/copyright-and-fair-use(lastvisitedJan.14,2017).

NiskanenCenter|12

Perhapsmostimportantly,underthecourts’currentstandardofevaluation,itisnotclearwhatis and is not infringement. The tests courts utilize vary from one circuit to the next, whichcreatesinherentuncertaintyabouthowacopyrightcasewillberesolved.Inaddition,thetestshavebeenappliedinconsistently,increasinguncertainty.45Toillustratethis,itmaybehelpfultoexplainwhatinfringementisanddescribethecurrenttestsconductedbythecourts.CopyrightInfringementCopyright infringement occurswhen any of the copyright owner’s exclusive rights under theCopyrightActhavebeenviolated.46Ifacopyrightownerbelievesaninfringementhasoccurred,theownermay sue thealleged infringer toenforce the rights theownerbelieveshavebeenviolated.47Therearetworequirementsfordemonstratingcopyrightinfringement:(1)ownershipofavalidcopyrightbytheplaintiff,and(2)copyingbythedefendant.48Thefirstrequirement,ownership,is typically established through proof of registration with the U.S. Copyright Office.49 Thesecondrequirement,“copyingbythedefendant,”consistsoftwoseparateelements:“copyinginfact,”and“copyingasalegalproposition.”50Element one, “copying in fact,” may be proven with direct or indirect evidence.51 If directevidence of copying cannot be shown, copying in fact may be shown indirectly bydemonstratingthattheallegedinfringerhadaccesstotheearlierwork,andthetwoworks inquestionshare“probativesimilarity.”52Element two, “copying as a legal proposition,” also known as “substantial similarity,”53 isarguably themore difficult element to prove. The circuits employ different tests to evaluatesubstantial similarity, and, as Nimmer on Copyright notes,54 “[a]lthough it is clear that thedeterminationofsubstantialsimilaritypresentsanissueoffact,thecorrectprocedureforthatdetermination remains clouded.”55 The traditional test for substantial similarity is the“AudienceTest,”whichHaroldLloydCorp.v.Witwerstatesasthefollowing:

45See,e.g.,supranote38.46CatherinePalo,CopyrightInfringementLitigation,in77AM.JUR.TRIALS449§23(updatedFeb.2016).47Id.48MELVILLEB.NIMMER&DAVIDNIMMER,NIMMERONCOPYRIGHT§13.01(MatthewBender,rev.ed.2015).

49Id.50Id.

51Id.52Id.53Id.at13-9to-10.1.54NIMMERONCOPYRIGHTisconsideredoneoftheleadingtreatisesonU.S.copyrightlaw.NimmeronCopyright,

LEXISNEXISSTORE,https://store.lexisnexis.com/products/nimmer-on-copyright-skuusSku10441(lastvisitedMar.28,2017).55NIMMER&NIMMER,supranote48at§13.03(E)(3).

NiskanenCenter|13

Thequestionreally involved insuchcomparison istoascertaintheeffectofthealleged infringing [work]upon thepublic, that is,upon theaverage reasonableman. If an ordinary person who has recently read the story sits through thepresentation of the picture, if there had been literary piracy of the story, heshould detect that fact without any aid or suggestion or critical analysis byothers. The reaction of the public to the matter should be spontaneous andimmediate.56

Many Circuits have modified the Audience test. For instance, the Second Circuit’s test forsubstantial similarity turnson (1)whether thedefendant copied theplaintiff’swork (aquerybased on “expert analysis and dissection”), and (2) whether that copying “constitute[s] animproper appropriation” (a jury evaluation that “would strictly apply the audience test andprecludedissectionandexpertanalysis”).57TheNinthCircuitemploysatwo-prongsubstantialsimilaritytest,consistingof(1)anextrinsicobjectiveanalysisconductedbythecourt(limitedto“specificcriteriawhichcanbelistedandanalyzed” such as “analytic dissection and expert testimony”), and (2) an intrinsic subjectiveanalysisconductedbythejury(basedon“theresponseoftheordinaryreasonableperson”).58OthercircuitshaveadoptedtheSecondCircuittest,theNinthCircuittest,acombinationofthetwo approaches, or variations on the themes of each.59 Therefore, the circuit evaluating acopyright infringement claimdictateswhich test likely applies. Yet, evenwhenarmedwith adescriptionofthetestsutilized,theoutcomesofcasesarenotconsistent,increasingthelevelofuncertainty.RecentCasesChallengingTraditionalInterpretationsofInfringementClaimsOne area thatmay be experiencing significant upheaval ismusic copyright law. It was onceconsideredverydifficulttoproveamusiccopyrightinfringementclaimincourt.60Severalwell-knowncasesillustratethisdifficulty.OneexampleistheSecondCircuitcaseSellev.Gibb.61ThesuitallegedthatTheBeeGees’hitsongHowDeep IsYour Love infringedanearlierworkwrittenbya relativelyunknownartist.56HaroldLloydCorp.v.Witwer,65F.2d1,18(9thCir.1933);NIMMER&NIMMER,supranote48,at§13.03(E)(3).

57NIMMER&NIMMER,supranote48,at§13.03(E)(3).58Sid&MartyKrofftTelevisionProds.,Inc.v.McDonald’sCorp.,562F.2d1157,1163–64(9thCir.1977).59NIMMER&NIMMER,supranote48,at§13.03(E)(3).60See,e.g.,DebraPrestiBrent,TheSuccessfulMusicalCopyrightInfringementSuit:TheImpossibleDream,7U.MIAMIENT.&SPORTSL.REV.229,229(1990)(representingthehistoricalviewthatconsidered“[m]usicalcopyrightprotection...amisnomer,”andunderthetraditionalregime,“[a]plaintiffseekingtoprotecthispropertyinterestfindslittlesympathyfromthejudiciary.”);WilliamR.Coulson,They’rePlayingOurSong!ThePromiseandthePerilsofMusicCopyrightLitigation,13J.MARSHALLREV.INTELL.PROP.L.555,575(2014)(“So,docomposersever‘borrow’fromeachother?Ofcourse,ithappens.Examplesabound.Evenifitamountstolarceny,isiteasytoprove?Decidedlynot.”);seealsoFrancis,supranote41,at494;Lemley,supranote38,at741–42.61SeeSellev.Gibb,741F.2d896(7thCir.1984).

NiskanenCenter|14

Thetwosongsinquestionsharedthirtyidenticalnotesandmorethanfortyidenticalrhythmicpatterns, a level of similarity that an expert witness testified could not have occurredindependently.62 Despite these similarities, an expert witness’s testimony, a jury decision infavoroftheplaintiff,andoneofTheBeeGeesmixingupthetwosongsincourtduringthetrial,thecourtrefusedtosupportafindingofaccessorinfringement.63Anotherexampleisthe2003NinthCircuitcaseNewtonv.Diamond,whereanexactcopyofsixsecondsofNewton’ssoundrecordingwassampledandloopedtorepeatmorethanfortytimesinaBeastieBoyssong.64Inacasewhererepeatedcopyingwasabsolutelyproven,thecopyingofthreeidenticalprotectednoteswasfoundtobetoominimaltoconstituteinfringement.65NinthCircuitcaseSwirskyv.Careydemonstratesthedifficultyofsurvivingsummaryjudgmentwhen bringing a music infringement claim.66 Swirskywas originally dismissed; however, onappeal, thecourtreversedbecausetheplaintiff’sexpertwitness identifiedagreatnumberofsimilarities shared by the songs.67 Though the songs shared nearly identical structures andhighly similarbass lines, chordchanges,and tempos, thecasewasoriginallydismissedundersummaryjudgment,arguablyduetothedifficultyofapplyingthelaw.The outcome of the recent Blurred Lines case may have upset this and other past legalprecedent.68In2013,songwritersforthehitsongBlurredLinesfacedallegationsthatthesonginfringedMarvinGaye’sclassicGottoGiveItUp.69Despitetheabsenceofanyidenticalnotes,chord progressions, lyrics, or melodies, the court found for the Gaye estate.70 Many havesuggestedthattheBlurredLinesverdictmayhave increasedconcernsovertheriskof liabilityformusiccopyrightinfringement.71

62Seeid.at899–906.63See id. at 899–906; Selle v. Gibb 741 F.2d 896 (7th Cir. 1984), MUSIC COPYRIGHT INFRINGEMENT RESOURCE,http://mcir.usc.edu/cases/1980-1989/Pages/sellegibb.html(lastvisitedMar.17,2017).64SeeNewtonv.Diamond,388F.3d1189(9thCir.2003).65Seeid.at1190,1196–97.66SeeSwirskyv.Carey,376F.3d841(9thCir.2004),asamendedondenialofreh’g(Aug.24,2004).67Seeid.at845–46.68Foramoredetaileddiscussion,pleaseseeReginaZernay,Comment,CastingtheFirstStone:TheFutureofMusicCopyrightInfringementLawAfterBlurredLines,StaywithMe,andUptownFunk,20CHAP.L.REV.177(2017),http://www.chapmanlawreview.com/wp-content/uploads/2017/04/20-1_Zernay.pdf.69ComplaintforDeclaratoryReliefat4,Williamsv.BridgeportMusic,Inc.,No.CV13-06004-JAK(AGRx)(C.D.Cal.Aug.15,2013),2013WL4271752.70Williamsv.BridgeportMusic,Inc.,No.LACV13-06004JAK(AGRX),2014WL7877773,at*6(C.D.Cal.Oct.30,2014).71See,e.g.,AdamR.Bialek,CaliforniaJuryFinds“BlurredLines”Infringed“GotToGiveItUp”:Society’sMixedSignalsOnCopyingandIntellectualPropertyRights,27INTELL.PROP.&TECH.L.J.14,15(2015);BlakeBrittain,MusiciansMoreCarefulAfter‘BlurredLines’Case,BLOOMBERGBNA(Sept.17,2015),http://www.bna.com/musicians-careful-blurred-n17179936188/;KeithMurphy,7Reasonsthe‘BlurredLines’VerdictShouldHaveEveryoneSpooked,BET(lastvisitedFeb.10,2016),http://www.bet.com/music/photos/2015/03/7-reasons-the-blurred-lines-verdict-should-have-everyone-spooked.html;TimWu,Whythe“BlurredLines”CopyrightVerdictShouldBeThrownOut,THENEWYORKER(Mar.12,

NiskanenCenter|15

Similarly, it is not always clear what role fair use will play in a copyright infringementcontroversy. The law states that the “fair use of a copyrighted work, including such use byreproduction in copiesorphonorecordsorby anyothermeans specifiedby that section, forpurposes such as criticism, comment, news reporting, teaching (includingmultiple copies forclassroomuse),scholarship,orresearch,isnotaninfringementofcopyright.”72Thus,iftheuseof copyrightedmaterial in a new work is considered fair use, the use is not infringing. Thecourtsusefourfactorstodetermineiffairuseapplies:

(1) The purpose and character of the use, including whether such use is of acommercialnatureorisfornonprofiteducationalpurposes;

(2) Thenatureofthecopyrightedwork;(3) Theamountandsubstantialityoftheportionusedinrelationtothecopyrighted

workasawhole;and(4) Theeffectoftheuseuponthepotentialmarketfororvalueofthecopyrighted

work.73Caselawsuggestsitisnotalwaysclearwhenandhowfairusewillapply.Forinstance,inCariouv.Prince,RichardPrince,“awell-knownappropriationartist,”74wassuedforinfringementwhenhe created and exhibited art that incorporated copyrighted photographs taken byphotographer Patrick Cariou without his permission.75 On appeal, the court held that whilePrincedidnotexpressanyintentto“comment[]onCariou’sworkoronculture,”suchanintentis not required toqualify for fair use.76 Instead, the court stated that the “critical” inquiry is“howtheworkinquestionappearstothereasonableobserver,notsimplywhatanartistmightsay about a particular piece or body of work. Prince’s work could be transformative evenwithoutcommentingonCariou’sworkoronculture,andevenwithoutPrince’sstatedintentiontodoso.”77ThecourtfoundthatPrince’sworkswerea“transformativeuse,”since“intwenty-five of his artworks, Prince has not presented the same material as Cariou in a differentmanner,butinsteadhas‘add[ed]somethingnew’andpresentedimageswithafundamentallydifferentaesthetic.”78Fairusehasalsobeenfoundtoapplywhereimagesare“transformativelydifferentfromtheiroriginal expressive purpose,” even though the images themselves have been used “in their

2015),http://www.newyorker.com/culture/culture-desk/why-the-blurred-lines-copyright-verdict-should-be-thrown-out.7217U.S.C.§107(2011),https://www.copyright.gov/title17/92chap1.html#107.73Id.74Cariouv.Prince,714F.3d694,699(2dCir.2013)(explainingthat“[t]heTateGalleryhasdefinedappropriationartas‘themoreorlessdirecttakingoverintoaworkofartarealobjectorevenanexistingworkofart.’”).75Id.at698.76Id.at707.77Id.78Id.at708.

NiskanenCenter|16

entirety.”79 In Bill Graham Archives v. Dorling Kindersley Ltd., Dorling Kindersley (“DK”)publishedGratefulDead:TheIllustratedTrip(“IllustratedTrip”),”acoffeetablebook“intendedasaculturalhistoryoftheGratefulDead.”80PlaintiffBillGrahamArchives(“BGA”),“claim[ed]to own the copyright to seven images displayed in Illustrated Trip, which DK reproducedwithoutBGA’spermission . . . . in significantly reduced form.”81 The imageswerepicturesofconcertpostersadvertisingGratefulDeadshows,greatlyreducedinsize,andwereincludedinthe book next to several other related pictures. The court found that “even though thecopyrightedimagesarecopiedintheirentirety,thevisualimpactoftheirartisticexpressionissignificantly limited because of their reduced size.”82 Because the remaining fair use factors“weigh[ed] in favor of DK’s use,” the use of BGA’s images in Illustrated Trip qualified as fairuse.83ItislikelythattheroleoffairuseinonlineinfringementcaseswillalsochangefollowingLenzv.UniversalMusicCorp.84InLenz,PlaintiffStephanieLenzposteda29-secondvideoonYouTubefeaturing a baby dancingwhile a song by the artist formerly known as Prince played in thebackground.85 Universal Music, which represents Prince, sent YouTube a DMCA takedownnotification for the video, claiming it infringed the song’s copyright. In response, Lenz suedUniversal,allegingthatUniversal failedto“firstevaluatewhetherthecontentqualifiesasfairuse,”inviolationoftheDMCA’stakedownnoticeprocedure.86Thecourtheld“thatthe[DMCA]statute requires copyright holders to consider fair use before sending a takedownnotification.”87Basedonthisdecision,itislikelythatthefairuseassessmentrequirementwillbemorestrictlyenforcedbythecourts,andtheabsenceofafairuseassessmentmayserveasadefenseforallegedinfringerschallengingDMCAtakedowns.It is also worth noting that the defense of accidental or “innocent” infringement is notrecognizedbycourts.Accidentalor“innocent”infringement“occurswhensomeoneengagesininfringing activity not knowing that her conduct constitutes infringement—perhaps mostcommonlywhen she knowingly copies fromanother’sworkbut reasonablybelieves thathercopyingisnotinfringing.”88Importantly,“since1931,adefendant’smentalstatehasclearlynotbeenrelevant...tothequestionofliabilityfordirectcopyrightinfringement.AstheSupremeCourt stated that year, ‘[i]ntention to infringe is not essential under the Act.’ So innocentinfringers are just as liable as those who infringe knowingly or recklessly.”89 Thus, innocentinfringementisstillactionable,andonceinfringementhasbeenshown,theinfringerisliable.

79BillGrahamArchivesv.DorlingKindersleyLtd.,448F.3d605,613–614(2dCir.2006).80Id.at607.81Id.82Id.at613,615.83Id.84SeeLenzv.UniversalMusicCorp.,815F.3d1145(9thCir.2015),cert.denied,137S.Ct.416(2016).85Seeid.at1148.86Id.87Id.88R.AnthonyReese,InnocentInfringementinU.S.CopyrightLaw:AHistory,30COLUM.J.L&ARTS133(2007).89Id.(internalcitationsomitted).

NiskanenCenter|17

Theoutcomesintheseandothercopyrightinfringementcasessuggestthatinsomesituations,the courts’ standard of evaluation may lead to inconsistent results. Offering a reliable andrelativelypredictableautomatedsystemofevaluationmayhelpminimizesuchinconsistencies.

B. ConcernsRegardingRegulationofOnlineDistribution

Withthedevelopmentoftechnologyanditsroleinbothcreatinganddistributingcopyrightedcontent,regulationhasbeenalongstandingissue.90Oneformofregulationthat isaconstantconcern is government regulation of the Internet. In a 2015 report issued by the FederalCommunications Commission (“FCC”), the importance of maintaining an open Internet wasmentioned:

The open Internet drives the American economy and serves, every day, as acriticaltoolforAmerica’scitizenstoconductcommerce,communicate,educate,entertain,andengageintheworldaroundthem.ThebenefitsofanopenInternetare undisputed. But itmust remain open: open for commerce, innovation, andspeech; open for consumers and for the innovation created by applicationsdevelopers and content companies; andopen for expansionand investmentbyAmerica’s broadband providers. For over a decade, the Commission has beencommittedtoprotectingandpromotinganopenInternet.91

ManyoftherecentdiscussionssurroundinggovernmentregulationoftheInternetfocusontheDMCA’s safe harbor provisions and its notice and takedown procedure. The entertainmentindustry inparticularhasbeenquite vocal about reforming theDMCA so thatOSPsareheldmoreaccountableforhosting infringingcontent.92Notsurprisingly,contenthostsandrelatedparties have protested such changes, citing concerns that it would chill creativity and thecontinueddevelopmentoftheInternetasaninformation-sharingtool.Contentownersandtheirrepresentativesclaimthatstrongerrulesarenecessarytoprotecttherights of copyright owners.93 Some have suggested that the DMCA notice and takedownprocedurehasbecomeanantiquatedandineffectivemeansofpolicinginfringement.94Others

90See,e.g.,LawrenceLessig,CodeIsLaw:OnLibertyinCyberspace,HARVARDMAGAZINE(June1,2000),http://harvardmagazine.com/2000/01/code-is-law-html.91FCC,REPORTANDORDERONREMAND,DECLARATORYRULING,ANDORDERFCC15-24(FCCRcd.)(2015),http://transition.fcc.gov/Daily_Releases/Daily_Business/2015/db0312/FCC-15-24A1.pdf.92See,e.g.,MicahSingleton&BenPopper,TheMusicIndustryCranksUptheVolumeinItsFightAgainstYouTube,THEVERGE(June3,2016,12:48PM),http://www.theverge.com/2016/6/3/11852146/music-industry-fighting-youtube-dmca.93See,e.g.,IrvingAzoff,DearYouTube:AnOpenLetterfromIrvingAzoff,RECODE(May9,2016,6:00AM),http://www.recode.net/2016/5/9/11609494/irving-azoff-youtube-artists-streaming-music.94See,e.g.,BRUCEBOYDEN,THEFAILUREOFTHEDMCANOTICEANDTAKEDOWNSYSTEM:ATWENTIETHCENTURYSOLUTIONTOATWENTY-FIRSTCENTURYPROBLEM(2013),http://cpip.gmu.edu/wp-content/uploads/2013/08/Bruce-Boyden-The-Failure-of-the-DMCA-Notice-and-Takedown-System1.pdf.

NiskanenCenter|18

havestatedthatthecurrentsystem“unfairlyfavor[s]technologycompaniesandroguepiratesites”and“forcescreatorstopolicetheentireInternetforinstancesoftheft,placinganundueburdenontheseartists.”95In contrast, contenthostsand relatedpartiesdiscourage increasedgovernment involvement.Opponents of increased regulation argue that greater restrictions will interfere with thecontinueddevelopmentoftheInternetasaplatformforinnovation,freeexpression,thebroaddistribution of content, and creativity.96 Some have stated that the DMCA takedown noticeprocedurehas, infact,succeededin“strik[ing]abalancebetweenfacilitatingfreespeechandcreativitywhileprotecting the interestsof copyrightholders”and“fuel[ing] thecreationofaboomingdomestic Interneteconomythatwasworthnearly$1trillionor6percentofGDP in2014.”97Othershaveevenexpressedconcernaboutthemotivationbehindthemedia’seffortstoincreaseregulation,withatleastonecommentatorclaiming:“TheInternetwouldhavebeendestroyed...butthemediadidnotcare,justaslongastheygottheirroyalties.”98Whatneithersidecan ignoreare theeffectsofexistingoverregulationcausedby thecurrentsystem of self-policing. While the DMCA notice and takedown process has helped contentdistributionandcreationthriveonline,theremaybesomeoverregulationintheDMCA’sself-policingregime.Evidenceindicatesthattheself-policingmethodsemployedbycontentownersand OSPs interpret copyright law so broadly that an alarming number of parties have beenimproperly accused of infringement.99 One study found that “in one in twenty-five cases,targeted content did not match the identified infringed work, suggesting that 4.5 millionrequests in the entire six-month data setwere fundamentally flawed.” In addition, “another15% of the requests raised questions about whether they had sufficiently identified theallegedlyinfringedworkoftheallegedlyinfringingmaterial.”100WhilethissystemhasnotpreventedtheexplosivegrowthofonlinecontentcreationsincetheDMCA came into effect, it has had an impact—one that may become more of a pressing95AshleyCullins,MusicIndustryA-ListersCallonCongresstoReformCopyrightAct,THEHOLLYWOODREPORTER(Mar.31,2016,1:48PM),http://www.hollywoodreporter.com/thr-esq/music-industry-a-listers-call-879718.96KerrySheehan,ThisYearinU.S.CopyrightPolicy:2016inReview,ELECTRONICFRONTIERFOUNDATION(Dec.25,2016),https://www.eff.org/deeplinks/2016/12/year-us-copyright-policy-2016-review.97DMCAInfringementTakedownsWork,INTERNETASSOCIATION,http://blog.internetassociation.org/post/141914748038/dmca-infringement-takedowns-work(lastvisitedJan.16,2017).98MikeKonrad,CopyrightStrangulation,AMERICANTHINKER(Sept.14,2013),http://www.americanthinker.com/articles/2013/09/copyright_strangulation.html#ixzz2euJwfxHT.99See,e.g.,URBAN,KARAGANIS&SCHOFIELD,supranote8,at2(“Study2’squantitativeanalysisrevealeddeficienciesinnoticeandtakedownprocedures,especiallyautomatedrequests....”);JonGosnell,Comment,KeepingtheInternetFree:WhytheDMCA’sSafeHarborProvisionShouldBeExpandedtoHelpCurbOverregulationofContentbyRemovingtheFinancialBenefitwithRightandAbilitytoControlExclusion1,3(2015),https://ssrn.com/abstract=2559455(“Forcontenthostingsites,thereisatensionbetweengoingtoofarandnotgoingfarenoughinprotectingcopyrightholders’rights....dotoomuch,andinformationisneedlesslystifled.Thisleadstoabsurdoutcomeslikepresidentialcampaignadsgettingflaggeddaysbeforeanelection,orfootagefromaNASAroverlandingonMarsgettingtakendownduetoaclaimfromanewschannel.”).100Id.

NiskanenCenter|19

concern in the future. It isworth examining the source of this overregulation and the effectautomatedsystemshaveinmisapplyingthelaw.Suchanoutcomemeritsattention,since“[t]hecontrol over the shifting of costs from one actor to an innocent other is often stated as ajustification for legal intervention.”101 A revised automated system for general use thataccurately reflects the proper legal standardmay one day offer a promising solution to thismisapplicationofthelawandotherrelatedissues.

C. OverregulationThroughSelf-PolicingTheDMCAoutlinesaprocedureforcopyrightownerstoenforcetheirrights,whilestillofferingOSPsprotection fromsecondary liability forhosting infringingcontent, so longas the serviceprovider acts expeditiously to remove infringing content upon notification (among otherrequirements).102TheDMCAnoticeandtakedownprocedureconsistsofseveralsteps.First,asnotedinLenz,thecopyright ownermust determine if the alleged infringement qualifies as fair use.103 If, afterconducting this analysis, the copyright owner has a good faith belief that the allegedinfringementdoesnotqualifyasfairuse,thecopyrightownermaysendatakedownnoticetoanagentdesignatedbytheOSPtoreceivetakedownnotificationsontheOSP’sbehalf.104Uponreceiving this notification, theOSPmust act expeditiously to remove the allegedly infringingcontent and promptly notify the user who posted the content that the content has beenremoved.105 The user who posted the allegedly infringing content may reply to the OSP bysending a counter notification explainingwhy the content is not infringing.106 If the counternotificationisproperlysentandreceived,theOSPmustputthecontentbackupandpromptlynotifythecopyrightownerthatthecontent isbeingrepostedwithin10businessdays,unlessthecopyrightownernotifiestheOSPthatheorshehasfiledacourtactionagainsttheallegedinfringer.107Scouring the Internet for potentially infringing material is an arduous task, one for whichcopyright owners and their representatives (sometimes collectively referred to as“rightsholders”) may not have the resources to achieve. In response to this difficulty,automated systems or “software bots” were created to “crawl the web for infringingmaterial.”108

101DonaldJ.Kochan,Bubbles(or,SomeReflectionsontheBasicLawsofHumanRelations),FORDHAMENVTL.L.REV.133,159(2015).10217U.S.C.512(c)(2011),https://www.copyright.gov/title17/92chap5.html#512.103Lenzv.UniversalMusicCorp.,815F.3d1145(9thCir.2015),cert.denied,137S.Ct.416(2016).10417U.S.C.512,https://www.copyright.gov/title17/92chap5.html#512.105Id.106Id.107Id.108URBAN,KARAGANIS&SCHOFIELD,supranote8,at31.

NiskanenCenter|20

Theeffectof thiswasa floodofDMCAtakedownnotices.109Forexample,“[i]n2009,Googlereceived 4,275 notices. In 2012, it received 441,370 notices.”110 By early 2016, “Googlewasreceivingbetween17and21millionrequestsaweek.”111Tocombatthedramatic increaseinDMCAtakedownnotices, severalof the largerOSPs turned toautomatedsystems toprocesstheoverwhelmingnumberofrequests.112OSPsarealsoveryprotectiveoftheirDMCAsafeharborprotections.113SecondaryliabilityforcopyrightinfringementwouldexposeOSPstoincalculablepotentiallosses,asmillionsofusersutilizeOSPs, andeach individual infringement constitutes a separate action.114 In aproactiveattempt toavoidhosting infringingmaterial in the firstplace, sites likeYouTube,Vimeo,andothershaveadoptedAIRSstodetectinfringementuponuserupload.115InthecaseofYouTube,qualifiedusersmayselecttheactionYouTubetakes if itsContent IDsystemdetectspotentialinfringement—blockthevideo,monetizethevideo,ortrackthevideo’sviews.116Forinstanceswheretherightsholderandtheallegedinfringerbothwishtomonetizethevideo,thedisputemay be remedied through a number of options, including an escrow-styled arbitration anddisputemechanism.117 These AIRSs have, on net, been a positive contribution to the onlinedigital landscape, ameliorating the need for disputes to take on a legal character, largelythroughmonetizationandvoluntaryremediation.These sophisticated self-policing systems, however, appear to have incorporated parametersthat go beyond what the law requires.118 As a result, a great deal of content is now beingimproperlyflaggedasinfringing.119Insomecases,partiesmayhavewillfullyabusedtheDMCAnoticeandtakedownprocess.120109Id.at32.110Id.111Id.112Id.at53.113Id.at8.114Id.at130(“Currently,thestatutedirectscourtstocalculatestatutorydamagesbyapplyingasetamounttoeachworkinfringed.This‘perinfringedwork’provisiondidnotanticipatethenumbersofworksthatuserscanplaceontolargeonlineplatformsortheabsurdresults—requestsofbillionsoreventrillionsofdollarsindamages—thatcanensue.FearofsuchstaggeringliabilitycostscancauseOSPstomakeoverlyconservativedecisionsregardingnoticeandtakedowncompliance.ThisisespeciallytruewhenOSPsmakedecisionsthatcouldaffecttheireligibilityforthesafeharborinthefirstplace,becausethelossofthesafeharborcouldresultinliabilityforapotentiallylargenumberofinfringedworks.”).115Seeid.at53;AndrewFlanagan,VimeotoLaunchMusicCopyrightIDSystem,BILLBOARD(May21,2014,10:00AM),http://www.billboard.com/biz/articles/news/digital-and-mobile/6092241/vimeo-to-launch-music-copyright-id-system-exclusive.116HowContentIDWorks,supranote15.117MonetizationDuringContentIDDisputes,YOUTUBE,https://support.google.com/youtube/answer/7000961?hl=en(lastvisitedApr.23,2017).118See,e.g.,Gosnell,supranote99.119URBAN,KARAGANIS&SCHOFIELD,supranote8,at2–3.120See,e.g.,KevinCollier,MeetRightsCorp,theAnti-PiracyCompanyThatCanKickYouOfftheInternet,THEDAILYDOT(July30,2014,5:00AM),http://www.dailydot.com/layer8/rightscorp-copyright-disconnect-internet-20-fine/;MichellePoe,WirelessInternetRouterExposesProvoWomanToThreats,Lawsuit,KUTV(Apr.23,2014,9:23AM),https://web.archive.org/web/20140706070103/http:/www.kutv.com/news/features/gephardt/stories/vid_62.sht

NiskanenCenter|21

As already mentioned, more famous examples include Paramount’s mismatched takedownnotice claiming Anchorman: The Legend of Ron Burgundy infringed An Officer and aGentleman,121 NBC’s takedown of a political campaign video from President Obama “daysbefore a critical voter registration deadline,”122 Scripps News Service’s removal of a publicdomainvideopostedbyNASAonNASA’sownYouTubepage,123andtheuseofthetakedownprocessbyapoliticalcandidatetoshutdownanotherpoliticalorganization’sYouTubeaccounttwice.124 Lesser-known examples can be found throughout the Internet. One website evenfeatures“DMCAHorrorStories.”125IncidentsincludeSony’sremovalofapoliticalcommentaryonYouTubethatfeaturedexcerptsfromMartinLutherKing’s“IHaveaDream”speechbecausea Sony-copyrighted song also included excerpts from the speech,126 and an unknown artistusingYouTube’snotificationprocesstofalselyclaimownershipofmaterial inordertoreceiveadrevenuefromanotherparty’svideo.127

D. AutomatedCopyrightEnforcementAsaPotentialSolutionAn ACES may reduce the above-described problems of uncertain outcomes, potentialgovernmentoverregulation,andoverregulationintheprivatesector.Foranautomatedsystemtoassistwith the largestnumberof cases, thepartsof thecourts’infringementstestscurrentlyappliedinconsistentlymustbeadjustedforconsistentapplication.Theactofestablishingmoreconsistenttestsislikelytocreatemorecertaintyinandofitself.Insomecases,theautomatedsystemcouldoffercompleteresolution.Inmoredifficultcases,thesystem’s evaluation may play a lesser role by serving as additional. In either situation, theintroductionof anobjectiveevaluation arguably increases the likelihoodofmorepredictableresults.The risk of government overregulation may also be minimized. A properly designed systemshould regulate strictlywithin the confines of the law. If this is achieved, and the system isrecognizedasconsistentandreliable,itmayhelplimithowmuch,ifany,additionalgovernmentregulationisnecessary.

ml;TakedownHallofShame,ELECTRONICFRONTIERFOUNDATION,https://www.eff.org/takedowns(lastvisitedJan.16,2017).121URBAN,KARAGANIS&SCHOFIELD,supranote8,at91.122NBCIssuesTakedownonViralObamaAd,supranote33.123Higgins,supranote34.124AttempttoSilencethePoliticalSpeechatRightWingWatch,supranote35.125DMCAHorrorStories,FIGHTFORTHEFUTURE:TAKEDOWNABUSE,https://www.takedownabuse.org/stories/toc/(lastvisitedMar.30,2017).126SonyTakedownoverMartinLutherKingSpeech,FIGHTFORTHEFUTURE:TAKEDOWNABUSE,https://www.takedownabuse.org/stories/sony_vs_fight_for_the_future/(lastvisitedMar.30,2017).127BoyInABargainBin,FIGHTFORTHEFUTURE:TAKEDOWNABUSE,https://www.takedownabuse.org/stories/royalty_free_claim/(lastvisitedMar.30,2017).

NiskanenCenter|22

Overregulation in the private sector appears to be a consequence of systems designed toprotecttheinterestsoftheuserscommissioningthesoftware,ratherthantoreflecttheactuallegalstandard.Thoughwideningthenetmayhelprightsholderscatchmoreinfringingcontentandpreventserviceprovidersfromhostingasmuchofit,asdescribedabove,thisapproachhasalready led to the occasional flagging of non-infringing content. Offering an objectivealternativemayhelpclarifythelinebetweenwhatispermittedbythelawandwhatconstitutesextra precautionarymeasures. In themeantime, voluntary private sector solutions likeAIRSsandotherCRSsseemtobeworkingeffectively.

II.AutomatedCRSsCurrentlyinUsebythePrivateSectorOneproposedmethodofcreatingtheautomatedsystemdescribedinthispaperisthroughthegovernment’sadoptionofamodifiedversionofexistingsystems,sincecreatingsuchasystemfromthegroundupcouldtakeyears.Thediscussionthatfollowsprovidesabriefexplanationofhow such systems function and how these systems could be modified to function for thepurposesdescribedherein.

A. CRSsCurrentlyInUseSeveral examples demonstrate how automated systems are currently being used to identifyinfringement.In addition to theelectronic systemsusedby schools todetect studentplagiarism,128writershaveaccesstoanumberoffreesoftwareapplicationsthatsearchforonlineplagiarismoftheirwork.129Online file-sharingwebsiteDropboxusessimilar software todigitally fingerprint filesand block potentially infringing files before upload.130 YouTube utilizes its own automatedContent ID system to evaluate content for infringement.131 Many other well-knownentertainment companies like Warner Bros, Sony, Disney, and even Facebook licenseautomatedcontentrecognitionsoftwarefromacompanycalledAudibleMagic.132NotealsothatYouTube’smonetizationfeatureintroducedanewmethodofhandlinginfringingcontent. Rather than automatically removing infringing videos, YouTube’s Content ID system128ProfessorsUseTechnologytoFightStudentCheating,U.S.NEWS&WORLDREPORT(Oct.3,2008,5:26PM),https://www.usnews.com/education/articles/2008/10/03/professors-use-technology-to-fight-student-cheating.129SeeOleksiySynelnychenko,HowtoProtecttheCopyrightofMyWebContent,IPWatchdog(June5,2013),http://www.ipwatchdog.com/2013/06/05/how-to-protect-the-copyright-of-my-web-content/id=40655/(providingalistoftoolstodetectonlineplagiarism).130SebastianAnthony,HowDropboxKnowsYou’reaDirtyPirate,andWhyYouShouldn’tUseCloudStoragetoShareCopyrightedFiles,EXTREMETECH(Mar.31,2014,2:07PM),https://www.extremetech.com/computing/179495-how-dropbox-knows-youre-a-dirty-pirate-and-why-you-shouldnt-use-cloud-storage-to-share-copyrighted-files.131HowContentIDWorks,supranote15.132AudibleMagicAccusesYouTubeofFraudOverContentIDTrademark,TORRENTFREAK(Jan.11,2017),https://torrentfreak.com/audible-magic-accuses-youtube-of-fraud-over-content-id-trademark-170111/.

NiskanenCenter|23

allowsqualifieduserstoinstructYouTubetorunadsoninfringingvideossorightsholdersmayautomaticallyreceiveadrevenue.133This type of monetization may have spurred an increase in content creation. Since firstlaunching its Content ID CRS, YouTube “has paid out $2 billion to rightsholders who havechosen to monetize claims” (as of July 2016).134 A corresponding increase in rightsholdersprofiting from the monetization of their works and content generation overall has beenobserved.Over8,000YouTubepartners,including“majornetworkbroadcasters,moviestudiosandrecordlabels”haveclaimed“over400millionvideos.”135Inaddition,“[a]sofMarch2015,creators filming inYouTubeSpaceshaveproducedover10,000videoswhichhavegeneratedover 1 billion views and 70+ million hours of watchtime.”136 These numbers suggest thatYouTube’s monetization CRS, a private sector AIRS, may be facilitating the progress of theusefularts.Aspopularascontentrecognitionsoftwarehasbecome,thecompaniessellingorusing itaretight-lippedaboutexactlyhowthesoftwareworks.137Duetothelimitedinformationavailable,what follows isageneraldescriptionofhowexistingcontentrecognitionsoftwareworksandhowitcouldbemodifiedtoservethepurposesrecommendedinthispaper.

B. HowAutomatedCRSsFunctionBased on the descriptions that are available, CRSs appear to have the following steps incommon:

(1) Thesubmissionofcopyrightedmaterialforinclusioninthesoftware’sdatabase;(2) The use of a software algorithm to analyze the material for distinctive features and

createa“digitalfingerprint”ofthematerialbasedonthisanalysis;(3) Theadditionofthedigitallyfingerprintedmaterialtothesoftware’sdatabase;(4) Theevaluationofpotentiallyinfringingmaterialbyscanningthenewmaterialwiththe

softwareandcomparingittothematerialinthesoftware’sdatabase;and(5) The flaggingofanymatchingmaterialbasedon theparametersprogrammed into the

software.138Slightly restated, an automated system functions by analyzing the work in question andcreating a digital “signature” or “fingerprint” of thework.139 The system then compares this133HowContentIDWorks,supranote15.134Id.135Id.136Statistics,YOUTUBE,https://www.youtube.com/yt/press/statistics.html(lastvisitedApr.25,2017).137URBAN,KARAGANIS&SCHOFIELD,supranote8at3(“Rightsholdersexpressedconcernthatpublicizingdetailsoftheirenforcementpracticesmightenablepiratestocircumventthem.OSPsworriedthatrevealingtheirpracticescouldsubjectthemtonegativeattentionfromrightsholders,targets,orotherOSPs,forcingthemtochangetheirpractices.”).138Strickland,supranote27.

NiskanenCenter|24

digitalfingerprinttoanexistingdatabaseofpreviouslyfingerprintedworkstoseeifamatchcanbe found.140 If thework in questionmatches an existingworkwithin the database, the newworkisflaggedasinfringing.141Thus, inordertofunctionoptimally,anautomatedsystemmusthaveanexistingdatabaseofdigitally fingerprinted material against which it can compare potentially infringing works.142Thereareanumberofways inwhichautomated systems candigitally fingerprintworks. Forinstance,thedigitalfingerprintingofmusicmaybeaccomplishedthroughseveralmeans:

Some programs analyze the tempo and beat of a song. Others measure thesong’s amplitude and frequency. Fingerprinting software usually takes severalsamples that last just a few seconds each from a single recording. A fewcompanies offer software that analyzes entire audio clips in order to get ascompleteafingerprintaspossible.Atleastonecurrentproductanalyzesasongfor landmarks -- distinctive acoustic moments in the clip -- then analyzes thesoundaround the landmarks. Ideally, the landmarkswill be readily identifiablewhenscanningothermusic.143

Oncethedatabasehasbeencreated,theautomatedsystemusesanalgorithmtodetermineifanewworkmatchesanexistingworkinthedatabase“withinacertainrangeofprobability.”144The comparison process has been likened to “the way forensics experts once matched asuspect’sfingerprintstothosefoundatacrimescene”inthat“[b]eforesophisticatedcomputersoftware and advancedmethods for examining fingerprints became available, expertswouldlookforpointsofsimilaritybetweendifferentfingerprints. Inmostcases,thespecialistwouldneedtodemonstrateat least16pointsofsimilarity foraprint tobeconsideredamatch.”145Likewise, content recognition software often allows users to indicate howmuch of amatchmustbepresentinorderforaworktobeflaggedasinfringing.146Whatthisprocesssuggests is thatarichdatabaseofexistingmaterial iscrucial totheproperfunctioningofanautomatedsystem.Additionally,theaccuracyofthedigitalfingerprintcanbeincreased based on the depth of the automated system’s analysis of a work. Finally, thereappearstobesomeflexibilityinwhatfeaturesanautomatedsystemwillrecognizeasamatch,andthethresholdatwhichasystemwillflagaworkforinfringement.Asa result,whileautomatedsystemsmaycurrentlybeprogrammedtosuit theneedsof theprivatesectorclientscommissioningtheiruse,itispossiblethatthesystemscouldbemodified139Id.140Seeid.141Seeid.142Seeid.143Id.144Id.145Id.146Seeid.

NiskanenCenter|25

tomatchthelegalstandardforinfringementmoreclosely.Suchamodificationcouldcreateatool foruseby thecourtsand thegeneralpublic to identifyactionable infringement.Even incaseswhere the system’s evaluation is not conclusive and human evaluation is required, anautomated system can provide useful evidence that encourages amore fair and predictableoutcome.

C. TheLimitationsandDrawbacksofAutomatedCRSsThe Limits of Filtering responds to calls by many “policymakers and copyright industrylobbyists”advocatingarequirementthatonlineserviceproviders implementCRSs inordertoreceiveDMCA safe harbor protections.147 The report engages in a detailed discussion of thecurrent limits and drawbacks of CRSs, ultimately concluding that “functionality and inherentlimitationsofthemostcommonfilteringtechnologies[]demonstratewhyamandatoryfilteringregimewouldposegravedangerstotheviabilityoftheinternetecosysteminexchangefor[]minimaleffectsononlineinfringement.”148Althoughthereportcorrectlynotes“suchtechnologiesarenotsufficienttoconsistentlyidentifyinfringementswithaccuracy,”itisimportanttorecognizethatatsomepoint—perhapssoonerrather than later—it may be possible to more closely align objective legal standards ofcopyright infringement evaluation with more robust and accurate CRSs.149 Thus, the nextsectionswillfocusonwhatsuchasystem,ifitweretomaterialize,mightlooklikeandhowitmightcomeabout.

III.PotentialUsesforanACESTheuseofautomatedsystemsfor infringementevaluationcouldsolveanumberofproblemswiththecurrentcopyrightenforcementsystem.Ownersofexistingcopyrightswouldbebetterequipped to protect their ownership rights, as automated systems could affordably identifyinfringementandencourageout-of-courtsettlementswithoutcostlylegalbattles.Newcontentcreators could avoid infringing before releasing material, leading to more certainty andprotectionfortheircopyrights.Theidentificationofpotentialinfringementmayevenfacilitatethelicensingofexistingcopyrightedmaterialandthecreationofinnovativenewworks.While the development of ACESs may be a task best suited for private organizations, thegovernmentcanplayasignificantrolebyencouragingdevelopmentofthesesystemsandcourtrecognitionoftheirvalidity.Wemayalreadybeheadinginthisdirection,asatleastonecourthas suggested automationmay serve as a legitimatemeans of fair use evaluation forDMCA

147ENGSTROM&FEAMSTER,supranote18,ati.148Id.149Id.at21(emphasisinoriginal).

NiskanenCenter|26

takedown notices.150 Expanding the role of automation to include infringement evaluationsdesignedforuseincourtscouldbethenextstep.OnepossibleuseofanACESisforprivateindividualstodetermineifinfringementhasoccurred.Forcontentcreators,thismayprovideamorereliablemethodtodetermineifoneoftheirnewworkshasaccidentally infringedonanexistingwork. If actionable infringement is found, thecontentcreatormayeithercontactthecopyrightownertonegotiatea license,ormodifythenewwork so that it is no longer infringing. Both solutions help reduce potential liability foraccidentalinfringement.For copyright owners, anACES that reflects the current legal standard couldprovide amoreaccurate assessment of whether another content creator has committed actionableinfringement.If infringementhasbeenfoundbythesystem,thecopyrightownermaybeinabetterpositiontocontacttheallegedinfringerandnegotiatealicensewithoutfilingsuit.Ifanagreementcannotbereached,theautomatedsystem’sfindingsmayplacethecopyrightownerinastrongerlegalposition.Forbothcopyrightownersandcreatorsofnewcontent,aproperly functioningACESoffersamore certain and protectable set of property rights. A finding of actionable infringementgeneratedbytheautomatedsystemgivesacopyrightownerstrongerevidencewithwhichtobring suit.Conversely, theabsenceofan infringement findingby theautomatedsystemmayprovideastrongdefenseagainstaninfringementclaim.AnACESmayalsofacilitatethelicensingofexistingmaterialbynewcontentcreators.Makingiteasier to find copyright owners and negotiate licenses could encourage progress andinnovation, and allow copyright owners tomaximize the value of their intellectual propertyrights.151Awell-designedACEScouldalsoserveasausefultoolforthecourts.Ifprogrammedcorrectly,suchasystemmayprovideamoreobjectivestandardofevaluation.Tofunctioncorrectly,theautomatedsystemmustbeprogrammedtomeetthecurrentstandardutilizedbytherelevantcourt fora findingof infringement. Itmayalsobenecessaryto identify inconsistencies inthecourts’applicationsofitstestsandadjustthesetoachieveamoreconsistentapplication.Oncethesystemhasbeenprogrammed,intheory,itshouldoperateinthesamemannereachtimeit

150SeeLenzv.UniversalMusicCorp.,801F.3d1126,1135(9thCir.2015),opinionamendedandsupersededondenialofreh'g,815F.3d1145(9thCir.2016),cert.denied,137S.Ct.416,196L.Ed.2d293(2016)(“considerationoffairusemaybesufficientifcopyrightholdersutilizecomputerprogramsthatautomaticallyidentifyfortakedownnotificationscontent....”).Note,however,thatonreview,theappellatecourtremovedreferencestotheuseofautomatedsystemsinitsamendedopinion.AshleyCullins,“DancingBaby”AppealsCourtDecisionStandsMinusthe“FairUse”Algorithms,THEHOLLYWOODREPORTER(Mar.17,20162:20PM),http://www.hollywoodreporter.com/thr-esq/dancing-baby-appeals-court-decision-876557.151THEDEPARTMENTOFCOMMERCE’SINTERNETPOLICYTASKFORCE,GREENPAPERONCOPYRIGHTPOLICY,CREATIVITY,ANDINNOVATIONINTHEDIGITALECONOMY,87–89(2013)[hereinafterGREENPAPERONCOPYRIGHTPOLICY],https://www.uspto.gov/sites/default/files/news/publications/copyrightgreenpaper.pdf.

NiskanenCenter|27

makes a determination, leading to more consistent and predictable results. As previouslymentioned,anobjectiveevaluationofwhetherinfringementhasoccurredisoneoftheprongsofboththeSecondandNinthCircuitsubstantialsimilaritytests.It is important to remember that some copyright controversies may not be solved withouthumanevaluation. Inthesecases,anautomatedsystemcannotandshouldnotbeviewedasthesolemeansofevaluation.Butneithershould itbeentirelydiscounted,as itmayserveasuseful,objectiveevidence.Keeping all this inmind, anACES, if programmed to function consistently and reliably, couldassistcourtsinmakingitsobjectivedeterminations.Thismaydiminishthecostsassociatedwithlong, drawn-out suits, and could be utilized to more fairly dismiss cases under summaryjudgment,or,whenappropriate,tomoveacaseforwardtoajury.

IV.PathwaystoPotentialImplementationA number of hurdles will need to be overcome before an effective automated copyrightenforcementsystemcouldbeactualized.Evenifsuchasystemwerewithinreach,thereareanumber of issues thatmaymake its implementation not only difficult, but arguably unwise.Nonetheless,thisisatechnologythatwillcontinuetomature;andasitdoes,policymakerswillincreasinglybefacedwithcallstoreformthecurrentcopyrightenforcementsystem.Undertherightpoliticalcircumstances,anACESmayendupbeingapolicyoptionworthconsidering,withsomeimportantcaveats.In the interest of considering a hypothetical future scenario, the discussion that followsdescribestheprerequisitesfortheemergenceofsuchasystem.

A.CreatingtheSoftwareSeveral significant challenges stand in theway of using existing software for the automatedsystemdescribedhere.Whiletheprivatesectorhasdevelopedautomatedsystemstodetectcopyrightinfringement,asdescribed above, many of the systems do not use an entire digital fingerprint to conductcomparisons,preventingacompletecomparisonofsongs.Aggravatingtheproblem,thereareseveralInternetsitesthatprovide“workarounds”toavoidgettingflaggedforinfringementbyautomated systems. One site recommends simply changing the pitch of an audio sample.152Anothersuggestsvariousalterationstoanaudio file, suchaspitch, timing,and length,which

152Ted,HowtoGetAroundThatPeskyCopyrighted-AudioFilteronYouTubeandFacebook,TURKEYMONKEY(July19,2009),http://www.turkeymonkey.com/2009/07/19/how-to-get-around-that-pesky-copyrighted-audio-filter-on-youtube-and-facebook/.

NiskanenCenter|28

can “thwart” YouTube’s Content ID system.153 At least one YouTube user has posted videoinstructionssuggestingthataddingadifferentsongtothebeginningandendofasongyouwishtopostmayalsobypassYouTube’sContentIDsystem.154Thus,while itmaybepossibletoreprogramexistingsoftwaretorequiretheuseofcompletedigitalfingerprintsforcomparisons,thisapproachmaybetechnicallyinfeasibleatthepresenttime.In addition to the ease of bypassing existing systems, processing speed and storagerequirementsmaybelimitingfactors.In2011,theInternationalSocietyforMusicInformationRetrieval releasedEchoprint–AnOpenMusic IdentificationService, awhitepaperdescribingthe function of Echoprint, the open-source CRS software currently used by Spotify.155Echoprint’s database at the time consisted of over 30 million tracks.156 In terms of storagespace,“[t]heindextakesroughly5gigabytesofdiskspaceper100,000tracks,whilethestoragerequires15gigabytes.”157Regardingprocessingtime“[q]ueriescanbeexecuted inpractice in100millisecondsaftercomputingthecodestringforasignal,whichoncurrenthardwaretakesunderatenthofasecond.”158KeepinginmindthatEchoprintislimitedtoaudiofilesandonlyevaluatesaportionofinformationabouteachaudiofile,itisunknownhowmuchmorestoragespace and processing speed would be required to create and store more complete digitalfingerprints.Itisalsounclearwhereandhowthesystemwouldbehosted.Finally,whileadjusting thesystemso itmoreclosely reflects therelevant legal standardmayonedaybewithinthereachofsoftwaredevelopers,forthetimebeing,themanyvariedtypesof content thatwouldneed tobe accurately fingerprinted limits this technology’s efficacy inproviding accurate matches for content beyond audio and video. As The Limits of Filteringnotes,nofingerprintingtoolcurrentlyexiststoevaluatesoftware,encryptedfiles,architecturaldesigns,orhandmade itemssoldonline.159Even foraudioandvideo,matching isnotoriouslydifficult.Until these challenges have been overcome, the available software cannot be consideredsuitablefortheACESthispaperenvisions.

153ScottSmitelli,FunwithYouTube’sAudioContentIDSystem,SCOTTSMITELLI(Apr.19,2009),http://www.scottsmitelli.com/articles/youtube-audio-content-id.1544Cracksn’Tuts,HowtoBypassWMGYouTubeCopyrightAudioMP3MusicFilesEasy,YOUTUBE(Aug.3,2012),https://www.youtube.com/watch?v=Pm_cpeJyVX0.155SeeDANIELP.W.ELLISETAL.,ECHOPRINT-ANOPENMUSICIDENTIFICATIONSERVICE(2011),http://ismir2011.ismir.net/latebreaking/LB-7.pdf.TheEchoprintwhitepaperisdiscussedatlengthbyTheLimitsofFiltering.SeegenerallyENGSTROM&FEAMSTER,supranote18.156ELLISETAL.,supranote155,at2.157Id.158Id.159ENGSTROM&FEAMSTER,supranote18,at2,14.

NiskanenCenter|29

B.DevelopingaDigitalLibraryAsdiscussedabove,anACEScannot functionoptimallywithouta richdatabaseofaccuratelyfingerprinted digital works. Fortunately, such a database may be within reach. A merger ofworks that have alreadybeendigitally submitted to theCopyrightOffice and theworks thathavebeensubmitted toprivatesectordatabasessuchasYouTube’sContent IDsystemcouldserveasanexcellentstartingpoint.There may also be government support for the creation of this database. In July 2013, theDepartmentofCommerce’s InternetPolicy Task Force released itsGreenPaperonCopyrightPolicy, Creativity, and Innovation in the Digital Economy.160 In its report, the Task Forceemphasized the importance of maintaining an accurate database of ownership information,specificallyforthepurposesoflicensing:

The most basic prerequisite for obtaining licenses is reliable, up-to-dateinformationaboutwhoownswhatrights inwhatterritories.Usersneedtofindthe right holders from whom to obtain permission, and right holders or theirrepresentatives need to be contacted to determine terms of use. As onlinebusinesses seek licenses for large repertoiresofworks tobeoffered inmultiplecountriesinavarietyofformats,andasmultimediausesbecomemorecommon,theneedforcomprehensiveglobally-linkeddatabasesisgrowing.161

TheHouse Judiciary Committee appears to share this sentiment.House Judiciary CommitteeChairman Bob Goodlatte (R-Va.) and Ranking Member John Conyers, Jr. (D-Mich.) recentlyissued a press release describing the “first policy proposal to come out of the Committee’sreviewofU.S.Copyrightlaw,”whichstated:

Today, we are releasing our first policy proposal, which identifies reforms tomodernize the Copyright Office so that it canmeet the challenges of the 21stCentury.AmongthereformsinthisdocumentaregrantingtheCopyrightOfficeautonomywithrespecttotheLibraryofCongress,requiringtheCopyrightOfficetomaintainanup-to-datedigital,searchabledatabaseofallcopyrightedworksandassociatedcopyrightownershipinformation,andmanyothersreforms.162

These reports suggest that the government has recognized the value of a searchable digitaldatabaseofcopyrightedworksandintendstobuildthisdatabasetomeettheneedsofdigitalusers.

160GREENPAPERONCOPYRIGHTPOLICYsupranote151.161Id.at89.162PressRelease,HouseJudiciaryCommitteeChairmanBobGoodlatte,Goodlatte&ConyersReleaseFirstPolicyProposalofCopyrightReview(Dec.8,2016)(onfilewithauthor).

NiskanenCenter|30

C.ClarifyingtheStandardUtilizedbyCourtsforInfringementItisimportanttonotethattheinconsistentstandardsusedbycourtstoevaluateinfringementmaymakeitdifficulttoprogramconsistentparametersintoanACES.However,thisshouldnotoperate as a complete bar to the use of an automated system. Rather, themost consistentelements of the courts’ standards of evaluation should be identified, the automated systemshouldbeusedwiththeexpectationthat itwillnotcompletelyreplaceallhumanevaluation,and any clearly inconsistent standards should either be modified or, in the alternative,acknowledgedandtreatedasknownexclusionstotheautomatedsystem.In cases where a 100%match of infringement has been found, the reliability of an ACES isunlikelytobechallenged.Rather,theburdenmayshifttotheallegedinfringertodemonstratethat the use was authorized, the plaintiff does not have ownership rights, or the use waspermittedunderadifferentdefense.It is incaseswherea lesserpercentageofpotential infringementhasbeendetectedthat thevalidityofanautomatedsystemmaybequestioned.Thismayoccurinvaryingdegreesandfora variety of reasons. Thus, it seems prudent to examine some of the reasons why minordissimilaritiesmayoccurandadjusttheroleoftheautomatedsystem’sevaluationaccordingly.DissimilaritiesDuetoPurposefulAlterationtoAvoidDetectionAs previously mentioned, content may be modified for the specific purpose of bypassingdetectionbyautomatedsystems.Examplesinclude:alteringthepitchofasong;163changingthetempoandspeedofasong;164orlimitingsnippetsofcopyrightedvideocontenttolessthanafewsecondstoavoidgetting“red-flagged.”165Insuchcases, thecontent is likely infringingdespite theseminormodifications.A reasonablesolution may be to program the automated system so that it does not judge whether thecontent is infringing, but, rather, it reports what percentage of similarity between the twoworksexists.

163Ted,supra,note152.164Smitelli,supranote153.165StephenMcArthur,HowtoBeataYouTubeContentIDCopyrightClaim–WhatEveryGamerandMCNShouldKnow,GAMASUTRA(June24,2014,4:41PM),http://www.gamasutra.com/blogs/StephenMcArthur/20140624/219589/How_to_Beat_a_YouTube_ContentID_Copyright_Claim__What_every_Gamer_and_MCN_Should_Know.php.

NiskanenCenter|31

DissimilaritiesDuetoPartialAppropriationSome forms of appropriation include the utilization of small snippets of copyrighted content(e.g., sampling),166 or the copying and incorporation of copyrighted elements or themes tocreate new works (e.g., derivative works).167 Such appropriation may be difficult for anautomatedsystemto identify,or,conversely,toofrequentlymisidentified ifthethresholdforidentificationissettoolow.Some of this may be addressed through programming. Setting the automated system toidentifysimilaror identicalelements thatcomposeasignificantportionof theworkcouldbeconsideredenoughforacourttotakenoticeandforpartiestoconsideraction.Forexample,perhaps the automated system could be set to flag three seconds of the same chordprogressionormelodybetweentwosongs,or5percentofaworkofartthatmatches5percentofanotherwork.Insuchcases,thesimilaritieswouldbeidentifiedandnotedbythesoftware,butusersshouldanticipatetheadditionalrequirementofhumanevaluation.SimilaritiesinEffectWithoutIdenticalorSimilarElementsTherecentBlurredLinescaseintroducedanarguablynewtheoryofcopyrightinfringement.168Under thedistrictcourt’sholding, ithasbeensaid thatmusiccopyright infringementmaybefoundwhenthestyleandfeelofasong issimilartoapreviouswork,even intheabsenceofidenticalchordprogressions,melodies,lyrics,orothermusicalelements.169SuchastandardisunlikelytobeprogrammedeasilyintoanACES,or,ifprogrammable,mayresultinanumberoffalsepositives.Twosolutionsareavailableforthissituation.Thefirstistodisregardtheuseofanautomatedsystemofevaluationinsuchcases,asitislikelytobeoutsidethecapabilitiesofthesoftware.Thesecondistoalterthecourt’sstandardofevaluation,since,initscurrentstate,thistheoryofinfringementseemstobemoresubjectivethanobjective.Thisiscontrarytothesubstantialsimilarity tests usedby the courts,which requireboth anobjective and subjective findingofinfringement.

166“Sampling”isdefinedas“theactofusingasmallpartofarecording(suchasasong)aspartofanotherrecording.”Sampling,MERRIAM-WEBSTER,http://www.merriam-webster.com/dictionary/sampling(lastvisitedJan.14,2017).16717U.S.C.§101(2011),https://www.copyright.gov/title17/92chap1.html#101(“A‘derivativework’isaworkbasedupononeormorepreexistingworks,suchasatranslation,musicalarrangement,dramatization,fictionalization,motionpictureversion,soundrecording,artreproduction,abridgment,condensation,oranyotherforminwhichaworkmayberecast,transformed,oradapted.Aworkconsistingofeditorialrevisions,annotations,elaborations,orothermodifications,which,asawhole,representanoriginalworkofauthorship,isa‘derivativework.’”).168SeegenerallyZernay,supranote68.169Id.

NiskanenCenter|32

Itmaybetoomuchtoaskthecourtstoadjusttheirstandardofevaluation.Suchacourseofaction may also take years to achieve. Rather, it seems wiser to exclude the use of anautomated system in cases such as these, and limit its application to controversies involvingmoreeasilyprovenformsofinfringement.

D.TheRoleoftheCourtsAtthistime,courtshavenotfullyembracedautomationasatoolforinfringementevaluation.Recent rulings suggest courts may be acknowledging the potential role of automation, buttherestillappearstobesomehesitancy.Thedistrictcourt’sholdingforLenzv.UniversalMusicCorp. indicated automation could be used to perform the fair use analysis required prior tosendingaDMCAtakedownnotice.170However,onreview,theappellatecourt“cutoutseveralparagraphsconcerninghowautomationmightbeleaneduponbycompanieslikeUniversalthathavetoevaluatefairuseonamassscale.”171Thus,automatedinfringementanalysismaystillbeconsiderednewtothecourts.Thiscouldbeanobstacletoadmissibility.UndertheFederalRulesofEvidence:

The United States Supreme Court has declared that the trial court has agatekeepingobligationtodeterminewhethertheexplanativetheoryunderlyingevery expert witness’ testimony, regardless of whether based on scientific,technical or other specialized knowledge, is “reliable.” . . . Federal Rule ofEvidence 702was amended effectiveDecember 1, 2000 . . . requiring that thetrial judge before permitting an expert to testify determine that “(1) thetestimonyisbaseduponsufficientfactsordata,(2)thetestimonyistheproductofreliableprinciplesandmethods,and(3)thewitnesshasappliedtheprinciplesandmethodsreliablytothefactsofthecase.”Asrestyled,Rule702nowprovidesthattheexpertwitnessestestimonymustbeshowntobe“theproductofreliableprinciplesandmethods”,Rule702(c), and that itmustalso [have]been shownthat“theexperthasreliablyappliedtheprinciplesandmethodsforthefactsofthecase,”Rule702(d).172

UntilanalysesbyACESsarerecognizedascommonlyacceptedandreliableformsofevidence,admissibilityofsuchevidencemayrequireclosereviewbycourts.173

170Lenzv.UniversalMusicCorp.,801F.3d1126,1135(9thCir.2015),opinionamendedandsupersededondenialofreh'g,815F.3d1145(9thCir.2016),cert.denied,137S.Ct.416,196L.Ed.2d293(2016);Cullins,supranote150.171Id.172Reliability“Gatekeeping”UnderDaubert/Kumho/Rule702:HistoricalDevelopmentandAssessment,5HANDBOOKOFFED.EVID.§702:5(7thed.2016).173Theprocessofestablishingautomatedsystemsasagenerallyacceptedmethodofinfringementevaluationbycourtsrequiresadetailedandinvolveddiscussionofitsown.Forthepurposesofthispaper,itmaybeenoughsimplytostatethatgeneralacceptanceofanynewmethodisnotautomatic,anduntilgeneralacceptanceis

NiskanenCenter|33

Fortunately,thesearenot impossibleobstacles.So longasthesoftware isproperlydesigned,refined to ensure accurate performance, and frequently updated to encompass the largestamountofcopyrightedmaterial,overtime,automationislikelytobecomeagenerallyacceptedformofevaluationandadmissibilitywillbecomelessofanissue.

V.PotentialObstaclesInadditiontotheconcernsraisedabove,anumberofissuesmayneedtobeaddressedbeforeimplementinganACESbecomespoliticallyandeconomically feasible. In this final section,wewill discuss some of the potential difficulties that may interfere with immediateimplementation.

A.TheRiskofMisuseMuchhasbeensaidaboutthemisuseofthecurrentDMCAnoticeandtakedownprocess.Itiswise to acknowledge that a similar version of this behavior may appear if an ACES isimplemented.Rather than allowing this possibility to exclude the use of automated systems altogether,solutionshavebeenproposedbyothersforcurrentsystemsthatmaybeequallyapplicabletotheACESproposedhere.Areportonarecentsetofstudiescommentedthat:

Theincreaseduseofautomatedsystemsbylargerightsholders,aswellas[OSPs],raised questions of accuracy and due process. Though rightsholders and OSPsgenerallyusesomeaccuracycheckstoday,weidentifiedaclearneedforbettermechanismstochecktheaccuracyofalgorithms,moreconsistenthumanreview,and a willingness by both rightsholders and OSPs to develop the capacity toidentifyandrejectinappropriatetakedownrequests.174

Thesestudiesdonotseemtoindicatethatautomationshouldberuledoutaltogether.Instead,proposed solutions include ensuring the accuracy of algorithms, incorporating more humaninvolvement in the evaluation process, and providing a better systemof responding to falsepositives.These solutions seem equally applicable to the ACES discussed here. During the initialimplementation period, it will be crucial to evaluate the system’s performance closely andrefine its programming in response to issues that will undoubtedly arise. As copyright lawfluctuates,itwillbenecessarytoadjusttheautomatedsystemaccordingly.Therearesituations

achieved,thereliabilityofthemethodwilllikelyrequireacase-by-casereviewwhencourtsconsideritsadmissibility.174URBAN,KARAGANIS&SCHOFIELD,supranote8,at3.

NiskanenCenter|34

wherehumanevaluationcannotbereplacedbyanautomatedsystem,andinsuchcases,theautomatedsystemmustbetreatedasasecondarytool,orconsideredinapplicable.Remainingmindfulof the limitationsofanACESwillallowcourtsandthegeneralpublic tomaximize itsusefulnessandaccuracy.

B.SupportfromtheSenateandHouseJudiciaryCommitteesActually implementing such a system would likely be too expensive for the government tojustify building its own database and automated system from the ground up. It may bepreferable to license the use of existing CRSs and the richest database(s) of digitallyfingerprintedcontent,thenmodifythelicensedcomponentstomatchtheneedsofthecourtsand the general public. This would likely be more cost-effective and efficient than thealternativeofbuildinganentirelynewautomatedsystemandcorrespondingdigitaldatabase.The Department of Commerce’s Internet Policy Task ForceGreen Paper on Copyright Policy,Creativity,andInnovationintheDigitalEconomyexpressedtheTaskForce’ssupportoffundingfor a U.S. public database, stating “it is important to ensure that the Copyright Office hassufficientresources”todevelopthedatabase.175YouTube’sContent IDsystemappearstobeastrongcandidatefor licensing. Itsdatabasehasexistedsince2006andmayverywellbeupdatedfrequently,sinceYouTubeisunderconstantinternal and external scrutiny. YouTube describes its Content ID database as “the mostcomprehensive in theworld,”with“morethan50millionactivereference files.”176Asoneofthe largest repositories of online content, YouTube likely has a rich database of digitallyfingerprintedmaterial.AudibleMagicmay also be an excellent option, as it is the company behind the automatedcontentrecognitionsoftwareutilizedbyShazam,WarnerBros,Sony,Disney,andFacebook.177Additionally,from2006to2009,YouTubelicensedsoftwarefromAudibleMagicforitsContentIDsystem.178Otherequivalentoptionsmayexist,andareworthexploring.Whilelicensinganautomatedsystemandmodifyingittosuitthepurposesdescribedheremaybe less expensive than building the system from the groundup, itmay still be costly. TheseexpensesarelikelytorequireapprovalfromtheSenateandHouseJudiciaryCommittees,thus,thesupportoftheseCommitteesiscritical.The expense could be offset by the reduced burden on courts, speedier resolution ofinfringement cases through the use of the ACES, and perhaps even increased copyright

175GREENPAPERONCOPYRIGHTPOLICY,supranote,at151.176Statistics,YOUTUBE,https://www.youtube.com/yt/press/statistics.html(lastvisitedMar.18,2017).177AudibleMagicAccusesYouTubeofFraudoverContentIDTrademark,supranote132.178Id.

NiskanenCenter|35

registrationincome,asmorecontentcreatorswillrealizetheyhaveavaluableandmorereadilyenforceablesetofrights.

C.TechnicalLimitationsTheLimitsofFilteringralliesagainstrecentcallsforDMCAreformthatwouldrequireOSPstoimplementCRSsinordertoqualifyforDMCAsafeharborprotections:

Ultimately, conditioning access to the safe harbor on an OSP implementing afiltering technologywouldundermine the certainty that theDMCAhas createdwithout a commensurate impact on infringement. Policymakers should resistcalls to upend the DMCA on the illusory hopes that filtering technologies canreplace the commonsense copyright regime that has allowed the internet toflourish.179

WhilemanyoftheconcernsraisedinTheLimitsofFilteringprimarilyspeaktothisissue,thereareseverallimitationsthatalsoapplytotheautomatedsystemproposedhere.TheLimitsofFilteringwarnsthat“allfilteringtoolscanbeevadedthroughbasicmanipulationofthefile.”180Thoughfingerprintingsystemsmaybelesssusceptibletosuchmanipulation,theyare not impervious, as the encryption of files “stymies even these more sophisticatedfingerprintingtechniques.”181Itisalsoimportanttonotethatcopyingcontentisnotalwaysconsideredinfringement:

[E]venifitwerepossibletoconsistentlyidentifycontentaccuratelywithfilteringtechnologies, such technologies are not sufficient to consistently identifyinfringementswithaccuracy,astheycanonly indicatewhetherafile’scontentsmatchprotectedcontent,notwhetheraparticularuseofanidentifiedfile isaninfringementinlightofthecontextwithinwhichthemediawasbeingused.182

This touches on the fair use argument raised in Lenz v. UniversalMusic Corp.,183 and lendssupporttothesuggestionthatanautomatedsystemcannotandshouldnotcompletelyreplaceallhumanevaluationofcopyrightcontroversies.The Limits of Filtering also notes that OSPsmay not have access to a complete database ofexisting material.184 This is a significant problem under the mandatory filtering regime The

179

ENGSTROM&FEAMSTER,supranote18,atii.180Id.at17–18.181Id.at18.182Id.183SeeLenzv.UniversalMusicCorp.,815F.3d1145(9thCir.2015),cert.denied,137S.Ct.416(2016).184ENGSTROM&FEAMSTER,supranote18,at18–20.

NiskanenCenter|36

Limits of Filtering advocates against; an incomplete database decreases the accuracy of anOSP’s filtering system, subjecting the OSP to the potential loss of its DMCA safe harborprotections due to circumstances beyond its control. Theoretically, the ACES described herewould be created using well-populated, pre-existing databases, making this problem lesspronounced. However, it is an important issue, as any CRS requires a rich and completedatabaseofaccuratelyfingerprintedmaterialtofunctionoptimally.Finally,TheLimitsofFilteringhighlightsthefactthat“[a]sencodingandcompressionstandardscontinuetoevolve,thefingerprintingtechnologiesthemselvesmustalsodeveloptokeeppacewith standards.”185 The paper correctly observes that “standardsmust continue to evolve tosupporttheproliferationofdigitalcontentonthe Internet.”186 Inordertomaintainaccuracy,CRSsmustbefrequentlyupdatedinresponse.FailuretodosocreatesahighriskthatCRSswillfallbehindtechnologyandceasetoproducereliableevaluations.Unfortunately,overcomingtheseandother limitationsshiftsthelikelihoodof implementingagovernment-adopted ACES further into the future. However, by recognizing the benefits ofsuchasystemandidentifyingitasagoal,thegovernmentanditsrelatedagenciesmaybegintotake actions that encourage more rapid evolution of CRSs in the direction of achieving aneffectiveversionofanACESforgeneraluse.

VI.PolicyRecommendationsAsdiscussedinprevioussections,thecurrentenvironmentseemstobefunctioningwell.What followsarepolicyrecommendations for legislatorsandregulators lookingatpotentiallyusingthistechnologyasacopyrightmanagementtool,noworinthefuture.

A.RefrainfromConditioningSafeHarborProtectionsontheUseofCRSsAt this time, the current copyright regime, warts and all, has functioned in a reasonablyeffectivemanner.Contentcreationisatanall-timehighandintermediaryliabilityprotectionsaffordedby theDMCAarerobustenoughtobalancethe interestsofOSPsandrightsholders.Whiletherearecertainlyissuesthatstillneedtobeironedout,suchasthelegaluncertaintiessurroundingthevariousapproachestodeterminingwhatconstitutes infringement, thestatusquoisenablingandspurringcreativecreationsandinnovation.Thoughtheremaybemeansofimprovinguponthestatusquo,thelargestsingleconcernatthistimeisadeteriorationofthecurrentenvironment.Asaresult,itisimperativethatpolicymakersrefrainfromadvocatingforadditional conditionality set on the safe harbor provisions of the DMCA. In particular, safeharbor provisions for OSPs should not be conditioned on the use of CRSs, automated orotherwise.

185Id.at21.186Id.

NiskanenCenter|37

Becausesomanyofthefingerprintingtoolscurrentlyinuseforautomatedcopyrightdetectionpurposes are designed to operate on single media types, no one overarching solution forautomated copyright management is currently feasible. The technical limitations of filteringtechnologies, the easily circumvented nature of many CRSs, and the “content bloom”engendered by the proliferation of digital communications technologies are all significantreasonstorefrainfromconsideringrulesthatwouldmandateautomatedcontentfilteringasapreconditionforsafeharborprotectionsundertheDMCA.ThepreviouslycitedreportfromEnginedescribeshowsuchamandatewouldunderminethecertaintyoftheDMCA’ssafeharborprovisions:

A safe harbor is only useful for mitigating uncertainty if the prerequisites forobtaining itsprotectionsareclear,but it isdecidedlyunclearhowpolicymakerscould precisely define what sorts of filtering technologies would be deemedadequatetoclaimthesafeharbor.Any filteringrequirementwouldeitherhavetoendorseaparticulartechnology(andquicklybecomeoutdated)orestablisha“reasonableness” standard that would require clarification from courts before[OSPs]couldhaveanyconfidenceintheirprotectionundersafeharbor.187

That diminished certainty, in turn, would have imposed significant costs on the health andvitalityoftheInternetecosystem.The report goes on to note that, “[t]he negative impact a proactive filteringmandatewouldhaveon[OSPs]—andtheconcomitantdecreaseincreativeoutputfromthosewhorelyonthe[I]nternetasamediumofproductionanddistribution—significantlyoutweighsitsbenefits.”188Anditconcludes:

[I]f the DMCA’s safe harbor protections were predicated on an [OSP]implementing a “reasonable” content filtering system, the uncertaintysurrounding any [OSP’s] eligibility for the safe harborwould undermine one ofthe central purposes of the DMCA: providing “more certainty . . . in order to[allow OSPs to] attract the substantial investments necessary to continue theexpansionandupgradingoftheInternet.”189

Inshort,voluntarilyimplementedAIRSs,aspreviouslydiscussed,canbeavaluabletoolforOSPsseeking to address concernsover copyright infringement.Mandating their use, or theuseofotherCRSs isanunnecessaryandpotentiallyunwiseapproach to remedyingonlinecopyrightinfringement concerns. Given the fragile order that has emerged online in the post-DMCA

187ENGSTROM&FEAMSTER,supranote18,at28.188Id.at25.189Id.at30.

NiskanenCenter|38

Internet ecosystem, it would be better for policymakers to refrain from actions that couldpotentiallyunderminethecurrentstatusquo.

B.AvoidOne-Size-Fits-AllSolutionsThe Limits of Filtering highlights a troublesome shortcoming in current CRSs—most onlyevaluateoneortwoformsofcontent:

Because fingerprinting technologies rely on algorithms that process theunderlyingmediacontentofagivenfile,theyarenaturallyconstrainedtoasmallsubset of copyrightable content. For example, because an audio fingerprintingtool’s central algorithm examines, say, the frequency values in a song file, itcannotbeusedtoidentifycopyrightedphotographsorsoftwareprograms,whichcontainnoaudio frequencyvalues.Assuch, to filterall filesonagivensite,an[OSP]wouldneedtoobtainadifferentfingerprintingtoolforeachtypeofmediathatis(orcouldbe)hostedacrosstheentiresite.Consideringtheincrediblywidescope of copyrightable content, there aremany types of content for which nofingerprinting toolexists, suchasarchitecturaldesignsorhandmade itemssoldonEtsy.190

Thus, Echoprint andShazamonlyevaluateaudio.Anti-plagiarism softwareusedby schools islikelylimitedtothewrittenword.YouTube’sContentIDsystemreviewsaudioandvideo.And,asThe Limitsof Filteringhas stated,CRSsmaynot yetexist forother formsof copyrightablecontent.Therefore, a one-size-fits-all solution is not only inappropriate, it is currently non-existent.Whether the development of a government-adopted system should be stalled until an all-encompassingsystemiscreatedisdebatable.IfanextremelystableandreliableCRSemergedthat specializes in one form of copyrightable content, such as music, it may be worthconsidering its adoption forpublic usewhilewewait forCRSs todevelop forother formsofcontent.Perhapsthisapproachismoreeffective,asitallowsfortheadoptionofseveralhighlyspecializedCRSsovertime,ratherthansettlingforasingleCRSthatmayoperateasajack-of-alltrades,butamasterofnone.

C.FacilitateOpenAccesstoDatabasesofCopyrightedWorksWherelegislatorscanhaveapositive impact intheapplicationofautomatedCRSs,theireyesshould focus on promoting tools that can positively contribute to growing the ecosystem ofonlinecontentproduction.

190Id.at16.

NiskanenCenter|39

IncentivizingrightsholderstocontributeaccurateanduptodatecontentidentifiersenrichesanACES’s database, and a rich database is a key component to ensuring the efficiency andaccuracy of matching content. Additionally, mechanisms for preventing uploads of falsefingerprintswillbenecessarytoensurethedatabaseisabletoprovideaccuracyandcertaintyamonguserswhodependonitssmoothoperation.

D.BroaderCopyrightOfficeReformOvertime,anumberofbroadreformstotheCopyrightOfficemaybenecessarytooptimizetheperformanceofthecopyrightenforcementsystemingeneral.As noted at the start of this paper, the long termof copyright duration has created amuchhigherriskofinfringement.Thismaybedueinparttothelimitedabilityofcontentcreatorstocome up with new and innovative content. Some analysts attribute the lessened ability ofmodernmusicians tocomeupwithmoresophisticatedcompositions toa“decline inmusicalliteracy.”191Thelimitedsetofartisticcontentthatthegeneralpublicfindsappealingmayalsobeacontributingfactor.AsnotedinDarrellv.JoeMorrisMusicCo.,“[i]tmustberememberedthat,whilethereareanenormousnumberofpossiblepermutationsofthemusicalnotesofthescale,onlya fewarepleasing;andmuchfewerstill suit the infantiledemandsof thepopularear.”192Ashortertermofcopyrightdurationmayhelpaddresstheseissues,asmoreworkswillbeconvertedtothepublicdomainandbecomeavailableforcontentcreatorstoutilize.TheCopyrightOfficeappears tohavereformplans. Inarecent interviewwithTheHollywoodReporter, Michelle Lee, then-Under Secretary of Commerce for Intellectual Property andDirectoroftheUnitedStatesPatentandTrademarkOffice(USPTO)discussedcopyrightsystemreform:

Our copyright system is really designed to play a critical role to promote andincentivize the creators of artistic works, while also promoting theirdisseminationanduse.TheInternethassignificantlytransformedhowcopyrightpolicyandcopyrightworksarebeingdistributednowadays.Computerscanmakean identical copy instantly. There is a tension between promoting andincentivizing that innovation and permitting free dissemination and fairdissemination.Thetwohavetoworkintandeminorderforustohaveasystemthatworks.TheCopyrightActhadn’tbeenchangedforareallylongtime.Noonehadstudiedit since the Clinton Administration. This was the first comprehensive look atcopyrightissuesrelatedtothedigitaleconomy.TheUSPTOtookthelead,andtheNationalTelecommunicationsandInformationAdministrationbothworkedonit,

191CharlesCronin,IHearAmericaSuing:MusicCopyrightInfringementintheEraofElectronicSound,66HASTINGSL.J.1187,1206,1224(2015).192Darrellv.JoeMorrisMusicCo.,113F.2d80,80(2dCir.1940).

NiskanenCenter|40

and we came up with a white paper. It’s on remixes, first sale doctrine andstatutorydamages.Welookedateverythingandgotalotofstakeholderinput.Wethenproducedthewhitepaperthatfocusedonthosethreetopics....It’sjustthebeginningbecause,aswithanypieceoflegislation,there’sstilllotsofworktobedoneandalotofstakeholderdiscussionstobehadintermsofbestpractices and so forth. Of course, in the background, ChairmanGoodlatte andothers in the House are looking at pretty comprehensive reviews of copyrightissuesandthey’vebeenhavinghearings.193

There has even been discussion of creating a separate court exclusively to handle copyrightinfringementcontroversies.Ms.Leestated:

We also noted the benefits of a new forum to adjudicate small copyrightinfringementclaimsbelowacertaindollaramountnotasencumberedbydistrictcourt litigationwithall thecostsandall thediscoverytoprovideanalternativeavenue to adjudicate small [dollar] amount copyright infringement cases. SothosewereconcretelegislativerecommendationsthatwemadetoCongress.194

Finally, as content is more widely created and distributed, the application of the fair usedoctrinemayberefined.TheCariouv.Princecasehighlightsthecurrentuncertaintyabouttheroleofintentinafairuseanalysis.BothCariouandBillGrahamArchivesv.DorlingKindersleyLtd.raisequestionsaboutwhatconstitutessufficienttransformationtoqualifyasfairuse.Lenzv.UniversalMusicCorpwilllikelyincreasetheroleofthefairuseanalysisintheDMCAnoticeandtakedownprocess.

ConclusionUpdatingthecopyrightenforcementsystemhasbecomeanimportantpriority,asthecurrentsystem does not address the exponential growth of digital content creation and onlinedistribution tools. Furthermore, the systemof self-policing that has resulted from theDMCAnoticeandtakedownprocedurehasevolvedintoaformofoverregulationthatfailstoapplythelawcorrectly.Inlightoftheseissues,agovernment-adoptedACESmayofferanidealsolution.By licensinganexistingautomatedsystem,modifying it tomatchthecurrent legalstandards,andfactoringinhumansupervisionforthemoredifficult infringementcontroversies,anACESmayhelpmodernize the current copyright systemandmore fairly and accurately reflect thelaw.193AshleyCullins,PresidentObama'sIPAdvisorTalksReformEffortsandaNewAdministration,THEHOLLYWOODREPORTER(Dec.23,20167:30AM),http://www.hollywoodreporter.com/thr-esq/president-obamas-ip-advisor-talks-reform-efforts-a-new-administration-936404.194Id.

NiskanenCenter|41

In spiteof thesepotentialbenefits, the limitationsanddrawbacksofexistingCRSsmayplacetheachievementofagovernment-adoptedCRSinthedistantfuture.Manychallengesmustbeovercome before an effective ACES can be implemented. In light of the effectiveness of theDMCA’scurrentsafeharborprocedures,augmentedbytheproliferationofvoluntarily-adoptedAIRSs, forthetimebeing,maintainingthecurrentDMCAframeworkandpromotingCRSsandvoluntarysolutions insupportof this frameworkmaybethemostsoundmethodofensuringthecontinuedprogressoftheusefularts.Despitethepresenttechnologicalconfines,thetimetostartconsideringtheimplicationsofanautomatedsystemisnow,asthisgoalmayhelpinfluencetheevolutionofexistingsystemsandleadtoasystemthatbetterbalancestheinterestsofallinvolved.