胡佛研究所-私人数据公共监管（英）-2021-28正式版.ppt

A HOOVER INSTITUTION ESSAYPrivate Data/Public RegulationBARRY FRIEDMANAegis Series Paper No.2105IntroductionPolice collect informationthat is what they do,and they could not do it without help.No doubt from the beginning of policing,officers have relied upon private parties to aidtheir investigations.Tipsters tip,snitches are paid,and well-meaning denizens share theinformation they possess.At crime scenes,police collect the names of people who mayhave seen something and follow up on those leads.And for the most part this has been,and remains,unregulated.But there has been a sea change,brought upon us by technology,a change so dramatic ithas transformed policing itself.Increasingly,the information police collect is digital.Fewersearch warrants,more requests for orders to harvest metadata.Purchasing large pools ofprivate data from data brokers.Capturing location information in various ways.Tappinginto a network of private security cameras.And so on.Sometimes police collect the datathemselves.More often they gather it from third parties.They do so from volunteers,bypurchase,and by court order.1The digital nature of the data allows it to be acquired in bulk.Rather than just focusingon the target of a particular investigation,the government can gather the informationindiscriminately about large swaths of the population or even about all of us.2 This wasthe case with the National Security Agencys collection of our telephone metadata,but italso is the case at the local level by automated license plate readers.3The uses of this data are so broad that it is not just what we traditionally think of as the“police”who gather the data,but a variety of government entities involved in public safetyand law enforcement.Bulk data collection has allowed law enforcement to shift its focusfrom a reactive investigative stance to a proactive and deterrent one.The police long haveinvestigated particular cases,with suspects they had in mind or were trying to identify.4Now,government agencies utilize large databases to predict where crime will occur,or minethem for evidence of criminality.5 This occurs without a particular suspect in mind and isaimed more broadly at the entire population or a segment of it.No doubt this data collection has value,but precisely how much is an open question,and may depend in part on its use.For example,data collected from third parties,suchas genetic information,has helpedto solve cold cases.6 Facial recognition was vital toidentifying the insurrectionists of January 6,2021.7 These are investigative uses.Here,the2value is plain,but the extent of that value unknown.The utility of bulk data collectionfor predictive or deterrent purposes is far more uncertain.8 In truth there is far too littleassessment of the utility of these new techniques,for whatever purpose they are deployed.But one thing is clear:placing this much personal data in the hands of the government hasits costs.It endangers our personal security,and our sense of privacy.It threatens our rightto associate,including for political activity.It puts enormous power in the government tocontrol behavior.And too often the collection is disproportionately of people of color.Im going to make a novel argument that,if adopted,would allow for capturing the benefitsof data-driven policing,while mitigating its costs.My argument is that government agenciesthat engage in policing cannot collect digital data,particularly about individuals for whomthere is no suspicion of wrongdoing,without a sufficient regulatory scheme in place.Thus,if these practices are to continue,legislative bodies must step in and adopt regulation.Whats novel about my argument is that I frame it in constitutional terms.Im not justarguing that regulation is desirable as a normative matter.My precise claim is that asa matter of constitutional law,unauthorized and unregulated bulk digital collection ofsurveillance data simply may not occur.Absent such authorization and regulation,courtsshould invalidate such collection.More bluntly:with or without court intervention,government agencies involved in the policing function should cease immediately whatthey are doing until sufficient authorization and regulation is in place.I support my argument by showing that in many contexts,under a number of constitutionalamendments,when government seeks to collect personal information,courts impose a setof“requisites”before such collection is mandated or upheld.These requisites include thatthe collection of data is statutorily authorized,that it furthers a legitimate governmentpurpose,that collection is minimized to protect privacy and personal security,that accessto the data is safeguarded,and that judicial review is available to challenge the collection,retention,and use of such data.Note that virtually all of this is missing from todays collection of surveillance data bygovernment.That is what must change.I am not arguing government cannot collect thedata,but that before it does,authorization and regulation by a democratically accountablebody is essential.This puts the decision to collect the data,and minimal safeguards attendantthereto,in proper hands,rather than letting government policing officials simply decide forthemselves.Although I hope it has very basic normative appeal,my argument is a complex one witha number of moving pieces.Im not going to have the space to lay it all out here,butI am doing so in other work.9 In Part I,I will describe the phenomenon of widespreaddigital surveillance data collection that draws my attention.In Part II,I will sketch out myBarry Friedman Private Data/Public Regulation3argument,briefly but in full.Part III then will hone in on the body of constitutional lawcentral to my argument,the cases I describe just above.Part IV will all-too-briefly touch onrelated parts of the argument and conclude.II want to begin by briefly describing the developments that motivate the paper and tohighlight two particular aspects.First,although many have expressed alarm that privatecompanies are collecting and aggregating enormous amounts of data about us,governmentsare building similar databases on their own for law enforcement purposes.There may bebenefits to all this,but there surely are costs.Yet,second,the Constitution as presentlyconstruedparticularly the Fourth Amendmenthas proven of little value in regulating thepractice of data collection,retention,and use.In the absence of constitutional regulation,there has been some legislative intervention,but far less than is necessary to cover the fieldor bring government surveillance data collection under the control it requires.I.APolicing agencies at the national and local level are building vast databases to keep tabson us all.The FBI is constructing an enormous biometric database that will include facialimages,iris scans,voice and palm prints.10 New York is one of several cities with an extensiveDomain Awareness system,which collects and aggregates information from a networkthat includes over 9,000 cameras,500 license plate readers,and data from governmentdatabases.11 No longer content with using state-run and regulated DNA databases,localdepartments have begun creating their own versions,snatching DNA in any way they canget it:by surreptitiously nabbing beverage cups from people they suspect,to asking consentfrom victims or people they wish to clear in investigations,and then retaining it.12 Thepolice use technology like from the company Cellebrite to vacuum up thecontents of cellphones during routine searches,and hoard it for later use.13Not content with their own data collection efforts,however,policing agencies increasinglyare relying on private data sources to monitor or provide information about the public forthem.One of the largest is the CLEAR database maintained by Thomson Reuters,whichadvertises CLEAR as a“powerful public records technology”that“brings all key contenttogether to provide intelligent analytics in one environment.”14 That database compilesextensive information on peoples credit,employment,and so forth.A recent story told howICE was using CLEAR to trace undocumented individuals using utility usage information.15Thomson Reuters is not alone;another recent news story related how ICE apparently isswitching data vendors,moving over to the database run by LexisNexis.16Databases that track our location are particularly illustrative of the public-privatepartnerships in this space.Automated license plate readers(ALPRs)provide an excellentexample.ALPRs sit on police cars or other vehicles,or are mounted in fixed locations,andHoover Institution Stanford University4they suck in the license plates of motorists around them,geolocating where those cars are atthat moment.17 Originally the point of ALPRs was to compare the reads to a“hot list,”suchas to detect stolen vehicles.18 But law enforcement agencies soon decided there was value inretaining the reads in case they proved useful later in a criminal investigation.19Many departments now use ALPRs extensively,both to create huge databases of whereautomobiles have been when,and to geofence their communities so they know who iscoming and going.Axon,the leading U.S.seller of body cameras and Tasers,is releasinga new digital ALPR system that turns ordinary police car dash cams into powerful toolsthat can collect reads from traffic across three lanes,front and back,as police cars moveon patrol.20 The LAPD has a database of some 320 million license plates that it stores for atleast five years.21 Myrtle Beach,South Carolina,keeps tabs on visitors with ALPR camerascovering all exit and entrance points to the shore town.In 2019 those cameras capturedalmost 40 million reads.22But policing agencies have expanded their capabilities substantially by linking up withprivate vendors to create a network of tracking and tracing capability.The industry leaderin this space is Vigilant,owned now by Motorola Solutions,which advertises and makesavailable to policing agencies a database of over 500 billion stored geo-located records.Vigilants database is fed with license plate readers attached to repo trucks,but also garneredfrom policing agencies that turn their information over,often in return for access to thelarger pool of data.23I.BThese public-private partnerships may have real advantages for policing agencies and forpublic safety,but one pervasive problem in this area is the lack of any systemic attempt toidentify those benefits.Cops will offer up anecdotes about the time ALPR evidence helpedcrack a case,but we dont know if the ALPR was essential,or how valuable it proved,letalone how often that was the case.There are well-known examples about how relianceon third-party data helped crack notorious cold cases,such as that of the Golden StateKiller.24 But how often does this happen?If we are going to allow the government to createor rely on third parties to provide huge aggregated databases,it seems incumbent upongovernment to prove the value.I will return to this point below.For now,though,I want to focus on the harms from this sort of data aggregation,becausethey are what make the case for regulation.If benefits were uncertain but harms nonexistent,our only concern would be the potential squandering of government resources.The loudpushback we hear today against government collection and use of private data suggestsmuch more is at stake.First,there are the threats from errors in these databases,and though error is a small word,the danger here is enormous.These mistakes cause people to become law enforcementBarry Friedman Private Data/Public Regulation5involved,and the news at present is all too clear about the risks those police encounterspose.At best they are an intrusion on liberty and psychologically stressful;at worst theyend up in police shootings or other uses of force.Albert Florence was picked up for anoutstanding warrant on a fine that had been paid but erroneously recorded as open.“Hespent seven days in jail”and“was strip-searched twice”in that time.25 Robert Julian-BorchakWilliams suffered a similar false arrest,this time because a facial recognition algorithm(with a human supposedly in the loop)wrongly IDed him for shoplifting from a high-endboutique.Despite his denying the image was him he spent 30 hours in jail,and his wifehad to claim an“emergency”to his employer lest he risk losing his job.26 It is impossible toknow the magnitude of these“errors,”but one can reasonably estimate that there are tensif not hundreds of thousands of stale or erroneous warrants sitting in databases waiting toresult in a wrongful arrest.27Second,theres misuse of the databases,just the kind of thing youd expect when you leavehuge stores of personal information lying about.A legislative audit found that over half ofthe 11,000 law enforcement personnel who searched the Minnesota Department of PublicSafety driver database conducted searches that were“questionable.”28 All too often there arestories of policing personnel dipping into the databases to spy on people with whom theyhave personal issuesfrequently stalking intimate partners.29Third,and perhaps most important,allowing government to accumulate and have easyaccess to detailed dossiers on all of us threatens our personal and collective security.30There are always those eager to argue it cant happen here,but it canand too often ithas.For decades,as part of its COINTELPRO operation,the United States intelligencecommunity not only kept tabs on the Civil Rights and Womens Rights movements,butall too often attempted to intervene in ways that were insidious to democracy.31“Fusioncenters”federally supported state and regional intelligence-gathering hubs,often relyingon privately collected datarepeatedly have been caught spying on entirely lawful FirstAmendment activity.32 Even if we accepted the purity of government watchersas I do,forexample,of the National Security Agencys vast data grab revealed by Edward Snowdenitsinconceivable that we want government to have access to all this data without public debateand adequate regulation.Although this sort of surveillance can fall upon all of us,its undeniably the case that whenthe government determines to collect and use information against people,theres a goodchance those who suffer most will be Black and brown.33 Its not a surprise that RobertWilliams and Albert Florence were Black.34I.CThe simple fact is that most government surveillance data collection,by policing agenciesor via third parties,is profoundly under-or unregulated.The Constitution as presentlyinterpreted offers almost no protection to the sort of widespread data collection by theHoover Institution Stanford University6government.35 There is some legislation in this space,such as the federal ElectronicCommunications Privacy Act and various state and local laws,but most of it is alreadyoutdated,and in general what police and private data collectors do happens without anyregulation.As Benjamin Wittes writes,aptly,“most of this data is not plausibly prot