Automatic License Plate Readers: Legal Status and Policy Recommendations for Law Enforcement Use

Americans drive. According to one survey, 83 percent of U.S. adults drive a car at least several times a week. footnote1_j7du9af 1 See Megan Brenan, “83% of U.S. Adults Drive Frequently; Fewer Enjoy It a Lot,” Gallup, July 9, 2018, https://news.gallup.com/poll/236813/adults-drive-frequently-fewer-enjoy-lot.aspx. In jurisdictions with limited or no public transportation, driving may even rival cell phone use as a modern necessity. Cars connect people with work, love, school, prayer, and protest.

They also leave a data trail. Historically, it would have been virtually impossible for law enforcement to routinely surveil all drivers. However, with the growing use of automatic license plate readers (ALPRs), police can now receive alerts about a car’s movements in real time and review past movements at the touch of a button. ALPRs could prove valuable in police investigations and for non–law enforcement uses like helping government agencies to reduce traffic and curb environmental pollution. But legal and policy developments have failed to adequately address the risks posed by this highly invasive technology. footnote2_m2qnmz7 2 The focus of this report is law enforcement use of ALPR. Although there is an important discussion and analysis to be had regarding non–law enforcement applications, such as ALPR use to help institute congestion pricing or to automate toll collection, they are outside the scope of this report.

Recent events crystalize ongoing concerns. With Black Lives Matter demonstrations taking place across the United States in the wake of the George Floyd and Breonna Taylor murders, law enforcement agencies large and small are deploying their expansive surveillance arsenals to monitor protesters. For many agencies, those surveillance tools include ALPRs, which have heightened relevance in localities where people must drive to protests, or if protests themselves are occurring by car, as is increasingly happening during the ongoing Covid-19 pandemic. footnote3_bxr1ckb 3 See, e.g., Katie Schoolov, “As Protests over the Killing of George Floyd Continue, Here’s How Police Use Powerful Surveillance Tech to Track Them,” CNBC, June 18, 2020, https://www.cnbc.com/2020/06/18/heres-how-police-use-powerful-surveillance-tech-to-track-protestors.html; Benjamin Wofford, “The Genius of Protesting in Car Caravans,” Washingtonian, June 1, 2020, https://www.washingtonian.com/2020/06/01/the-genius-of-protesting-in-car-caravans/; Caroline Haskins and Ryan Mac, “Here Are the Minneapolis Police’s Tools to Identify Protesters,” BuzzFeed News, May 29, 2020, https://www.buzzfeednews.com/article/carolinehaskins1/george-floyd-protests-surveillance-technology; and Catherine E. Shoichet, “They Can’t March in the Streets. So They’re Protesting in Their Cars Instead,” CNN, April 14, 2020, https://www.cnn.com/2020/04/14/us/coronavirus-car-protests/index.html.

The pandemic adds an additional dimension for consideration, as states look for creative ways to control the virus’s spread. With car travel expected to increase as states begin slowly loosening restrictions, ALPRs may play a larger role in law enforcement. footnote4_j66eeq7 4 See Haixia Wang, “41 Percent of Americans Say First Trip Will Be by Car within 100 Miles: Skift Research Travel Tracker,” Skift, May 1, 2020, https://skift.com/2020/05/01/41-percent-of-americans-say-first-trip-will-be-by-car-within-100-miles-skift-research-travel-tracker. States such as Rhode Island have already directed law enforcement to look for New York license plates in order to identify people who should be directed to self-quarantine. footnote5_ngsdole 5 Prashant Gopal and Brian K. Sullivan, “Rhode Island Police to Hunt Down New Yorkers Seeking Refuge,” Bloomberg, March 27, 2020, https://www.bloomberg.com/news/articles/2020–03–27/rhode-island-police-to-hunt-down-new-yorkers-seeking-refuge. Law enforcement agencies may look to automate this process by using ALPR devices to alert officers any time an out-of-state license crosses into their localities.

This white paper explains how ALPR technology works, focusing on its use by law enforcement agencies. It then analyzes both the legal and policy landscapes, including how courts have ruled on the use of ALPRs, and how they can be expected to rule in the future. Next, it outlines a series of concerns, ranging from high error rates to the impact on civil liberties and civil rights. Finally, it concludes with a set of recommendations for law enforcement, lawmakers, and technology vendors to enhance transparency and accountability and mitigate the impact of this technology on individuals’ civil liberties and civil rights.

Automatic license plate readers use a combination of cameras and computer software to indiscriminately scan the license plates of every car passing by. The readers, which can be mounted on stationary poles, moving police cruisers, and even handheld devices, log the time and date of each scan, the vehicle’s GPS coordinates, and pictures of the car. Some versions can also snap pictures of a vehicle’s occupants and create unique vehicle IDs. footnote1_1b9724g 1 Catherine Crump, You Are Being Tracked: How License Plate Readers Are Being Used to Record Americans’ Movements, American Civil Liberties Union, New York, NY, July 2013, 4, https://www.aclu.org/files/assets/071613-aclu-alprreport-opt-v05.pdf; see also Dustin Slaughter, “Philly Police Admit They Disguised a Spy Truck as a Google Streetview Car,” Vice, May 12, 2016, https://www.vice.com/en_us/article/kb77dm/philly-police-admit-they-disguised-a-spy-truck-as-a-google-streetview-car; Sam Biddle, “Hacked Border Surveillance Firm Wants to Profile Drivers, Passengers, and Their ‘Likely Trip Purpose’ in New York City,” Intercept, July 9, 2019, https://theintercept.com/2019/07/09/surveillance-perceptics-new-york-city-drivers; and Lily Hay Newman, “New Traffic-Enforcement Tech Peers into Your Car and Counts Passengers,” Slate, April 28, 2015, http://www.slate.com/blogs/future_tense/2015/04/28/automated_vehicle_occupancy_detection_looks_in_cars_counts_passengers_records.html. The devices send the data to ALPR software, which can compare each plate against a designated “hot list.” Such lists can include stolen cars and cars associated with AMBER Alerts for abducted children. footnote2_ghamaed 2 See, e.g., Crump, You Are Being Tracked, 5; International Association of Chiefs of Police, Privacy Impact Assessment Report for the Utilization of License Plate Readers, Alexandria, VA, September 2009, 2, 24–26, https://web.archive.org/web/20131024095529/http://www.theiacp.org/LinkClick.aspx?fileticket=N%2BE2wvY%2F1QU%3D&tabid=87; and New York v. Davila, 901 N.Y.S.2d 787, 789 (2010) (describing use of a hot list). They can also reference vehicles that are listed in local and federal databases for reasons that may include unpaid parking tickets or inclusion in a gang database. footnote3_tmgmyli 3 See, e.g., “National Crime Information Center (NCIC),” Federal Bureau of Investigation, Washington, DC, accessed August 26, 2020, https://www.fbi.gov/services/cjis/ncic (showing that NCIC database includes 21 files, including a Gang File and an Immigration Violator File); see also David J. Roberts and Meghann Casanova, Automated License Plate Recognition (ALRP) Systems: Policy and Operational Guidance for Law Enforcement, International Association of Chiefs of Police, Alexandria, VA, September 2012, 6, 22, https://www.ncjrs.gov/pdffiles1/nij/grants/239604.pdf. These queries happen automatically, though officers can also query plates manually. footnote4_od3q2r9 4 Roberts and Casanova, Automated License Plate Recognition Systems, 10.

In addition to checking data in real time, many cities and agencies retain plate information for future use, sometimes indefinitely. footnote5_l457iw4 5 Roberts and Casanova, Automated License Plate Recognition Systems, 8–9. This data can be used to plot a particular vehicle’s various locations or to identify all the cars at a given location, and it can even be analyzed to predict routes and future locations of a vehicle or set of vehicles. footnote6_puzjw6u 6 Crump, You Are Being Tracked, 5–6; Dave Maass and Beryl Lipton, “Data Driven: What Is ALPR?” Electronic Frontier Foundation, San Francisco, CA, November 15, 2018, https://www.eff.org/pages/what-alpr. These tools may cost little or nothing for police, often because the drivers themselves shoulder the cost of the technology through a fee charged on top of traffic ticket costs. footnote7_xnxrsqw 7 Eric Markowitz, “Pay This Fee, or Go to Jail: How License Plate Scanner Vigilant Solutions Makes [sic] Money in Texas,” International Business Times, February 3, 2016, http://www.ibtimes.com/pay-fee-or-go-jail-how-license-plate-scanner-vigilant-solutions-makes-money-texas-2290835. Notably, drivers in some jurisdictions can be jailed for failure to pay the private company’s fee. footnote8_r3wbdgb 8 Markowitz, “Pay This Fee, or Go to Jail.”

Law enforcement use of ALPRs is rapidly expanding, with tens of thousands of readers in use throughout the United States; one survey indicates that in 2016 and 2017 alone, 173 law enforcement agencies collectively scanned 2.5 billion license plates. footnote9_zklcppp 9 See Tanvi Misra, “Who’s Tracking Your License Plate?” Bloomberg CityLab, December 6, 2018, https://www.citylab.com/equity/2018/12/automated-license-plate-readers-privacy-data-security-police/576904. According to the latest available numbers from the Department of Justice’s Bureau of Justice Statistics, 93 percent of police departments in cities with populations of 1 million or more use their own ALPR systems, some of which can scan nearly 2,000 license plates per minute. footnote10_pzeq6p6 10 Brian A. Reaves, Local Police Departments, 2013: Equipment and Technology, Bureau of Justice Statistics, U.S. Department of Justice, Washington, DC, July 2015, 4, https://www.bjs.gov/content/pub/pdf/lpd13et.pdf. In cities with populations of 100,000 or more, 75 percent of police departments use ALPR systems. footnote11_9nphzn4 11 Reaves, Local Police Departments, 4. In some of the largest U.S. cities, millions of license plates are scanned over the course of a year. footnote12_7e7hobq 12 See, e.g., Dave Maass and Beryl Lipton, “Data Driven: Explore How Cops Are Collecting and Sharing Our Travel Patterns Using Automated License Plate Readers,” Electronic Frontier Foundation, San Francisco, CA, November 15, 2018, https://www.eff.org/pages/automated-license-plate-reader-dataset (finding that between 2016 and 2017, the Los Angeles County Sheriff’s Department scanned 234.36 million license plates with a 0.22 percent hit rate, the San Bernardino County Sheriff’s Department scanned 162.69 million license plates with a 0.06 percent hit rate, and the Sacramento Police Department scanned 116.23 million license plates with a 0.1 percent hit rate). According to a 2020 California state auditor report, the Los Angeles Police Department (LAPD) alone has accumulated more than 320 million license plate scans, and the Sacramento Police Department recorded up to 1.7 million scans in just one week. footnote13_kayg1ly 13 Auditor of the State of California, Automated License Plate Readers: To Better Protect Individuals’ Privacy, Law Enforcement Must Increase Its Safeguards for the Data It Collects, Sacramento, CA, February 2020, 12, http://auditor.ca.gov/pdfs/reports/2019–118.pdf. Despite this expansive data collection effort, many departments have not developed a policy to govern the use of ALPR technology, or provided privacy protections. While states such as California and Nebraska have passed laws requiring their departments to establish ALPR policies, not all departments have complied. footnote14_i2o4u6i 14 See Cal. Civ. Code § 1798.29; Neb. Rev. Stat. § 60–3206; Auditor of the State of California, Automated License Plate Readers, 2 (finding that the LAPD “has not developed an ALPR policy at all”).

Law enforcement use of ALPR data is not limited to reads captured by departments’ own devices; many departments have contracts with vendors that grant them access to private databases containing scans from private ALPRs and from other local and federal law enforcement agencies. For example, Vigilant Solutions (owned by Motorola Solutions), a leading provider of ALPR data to police based in Livermore, California, sells access to its database of more than 5 billion license plate scans collected across the country, including 1.5 billion reads provided by law enforcement agencies. This process creates a revolving door of license plate scans from law enforcement to Vigilant Solutions back to law enforcement agencies. footnote15_cd5n2hh 15 See Vasudha Talla, “Documents Reveal ICE Using Driver Location Data from Local Police for Deportations,” American Civil Liberties Union, New York, NY, March 13, 2019, https://www.aclu.org/blog/immigrants-rights/ice-and-border-patrol-abuses/documents-reveal-ice-using-driver-location-data.

Moreover, access to ALPR tools and data is not limited to law enforcement. For example, government agencies use license plate readers to automate toll collection and for pollution research; businesses analyze ALPR location data when assessing loan applications to help verify an applicant’s listed home address or to detect commercial use of vehicles when analyzing insurance claims; and private individuals and neighborhood associations can buy ALPRs for home and neighborhood security purposes. footnote16_8kkb1b0 16 See “Electronic Toll Collections: Easy System Integration for Vehicle Recognition,” Perceptics, Farragut, TN, accessed August 26, 2020, https://www.perceptics.com/markets/electronic-toll-collection/; Alexis Rivas, “Automatic License Plate Scanners Get Green Light for Pollution Research in San Diego,” NBC San Diego, November 22, 2019, https://www.nbcsandiego.com/news/local/automatic-license-plate-scanners-approved-for-pollution-emissions-research-san-diego-barrio-logan-national-city/2178630/; “Solutions,” Digital Recognition Network, Fort Worth, TX, accessed August 26, 2020, https://drndata.com/solutions; and Sam Dean, “Neighbors Are Using These Smart Cameras to Track Strangers’ Cars — and Yours,” Los Angeles Times, September 13, 2019, https://www.latimes.com/business/story/2019–09–12/flock-safety-license-plate-readers-los-angeles. These private actors can maintain their own hot lists of flagged license plate numbers and can share any data they collect with law enforcement at their discretion. footnote17_9ukc15h 17 Dean, “Neighbors Are Using These Smart Cameras to Track Strangers’ Cars.” Similarly, public agencies that collect and store ALPR data for non–law enforcement purposes may hold onto a dataset that proves alluring for police departments.

The U.S. Constitution’s Fourth Amendment protects people from unreasonable searches and seizures. footnote1_8ek6dwn 1 State law may provide additional protections that go further than the guarantees of the Fourth Amendment. See, e.g., Electronic Communications Privacy Act, Cal. Penal Code § 1546. According to the U.S. Supreme Court, the Amendment’s purpose “is to safeguard the privacy and security of individuals against arbitrary invasions by government officials.” footnote2_44n21hu 2 Carpenter v. United States, 138 S. Ct. 2206, 2213 (2018) (quoting Camara v. Municipal Court of City and County of San Francisco, 387 U.S. 523, 528 (1967)). Until the late 1960s, the Supreme Court ruled that Fourth Amendment protections only applied to searches and seizures of tangible property. footnote3_9wkifex 3 See Olmstead v. United States, 277 U.S. 438, 464–66 (1928). But in 1967, the Court expanded Fourth Amendment protections, holding in Katz v. U.S. (1967) that “the Fourth Amendment protects people, not places.” footnote4_262ra8x 4 Katz v. United States, 389 U.S. 347, 351 (1967). Specifically, the government was now prohibited from intruding upon a person’s “reasonable expectation of privacy.” In other words, if an individual seeks to keep something private, and that expectation of privacy is “one that society is prepared to recognize as reasonable,” the Fourth Amendment is triggered, and the government generally must obtain a warrant supported by probable cause before conducting a search. footnote5_tqr372l 5 See Smith v. Maryland, 442 U.S. 735, 740 (1979) (quoting Katz, 389 U.S. at 361). This approach seeks to protect the “privacies of life” from “arbitrary power,” and to “place obstacles in the way of a too permeating police surveillance.” footnote6_9b55b7j 6 Carpenter, 138 S. Ct. at 2214 (quoting Boyd v. United States, 116 U.S. 616, 630 (1886); and United States v. Di Re, 332 U.S. 581, 595 (1948)).

By contrast, the Court has not required a warrant or other heightened standard for police officers to take pictures of individual license plates and compare them against a law enforcement database. Its reasoning has been twofold. First, due to “the pervasive regulation of vehicles capable of traveling on the public highways,” there is no expectation of privacy in the content of license plates. footnote7_ww9fg8y 7 California v. Carney, 471 U.S. 386, 392 (1985); see also New York v. Class, 475 U.S. 106, 113 (1986) (“Automobiles are justifiably the subject of pervasive regulation by the State. Every operator of a motor vehicle must expect that the State, in enforcing its regulations, will intrude to some extent upon that operator’s privacy.”). Second, longstanding precedent holds that drivers on public roads cannot expect their movements to be kept private from the police since they could be observed by any member of the public (though, as discussed below, this presumption is beginning to shift). In keeping with these doctrines, courts have regularly held that law enforcement officers may, at their discretion and without any suspicion of criminal activity, perform at least an initial check of a license plate against a law enforcement database. footnote8_il93dff 8 See U.S. v. Knotts, 460 U.S. 276, 281 (1983); see also Cardwell v. Lewis, 417 U.S. 583, 590 (1974) (plurality opinion); Town of Woodworth, 132 So.3d 422 (La. App. 2013); State v. Davis, 239 P.3d 1002 (Ore. App. 2010); State v. Myrick, 659 A.2d 976; Davila, 901 N.Y.S.2d 787; U.S. v. Diaz-Castaneda, 494 F.3d 1146 (9th Cir. 2007); U.S. v. Ellison, 462 F.3d 557 (6th Cir. 2006); Olabisiomotosho v. City of Houston, 185 F.3d 521, 529 (5th Cir. 1999); United States v. Walraven, 892 F.2d 972, 974 (10th Cir. 1989); and United States v. Matthews, 615 F.2d 1279 (10th Cir. 1980).

Even so, there have long been hints that the tracking of vehicles’ movements could, under some circumstances, trigger Fourth Amendment concerns. As far back as 1979, the Supreme Court declared that “an individual operating or traveling in an automobile does not lose all reasonable expectation of privacy simply because the automobile and its use are subject to government regulation.” footnote9_g6bc0th 9 Delaware v. Prouse, 400 U.S. 648, 662 (1979). Similarly, when the Court analyzed the use of beeper technology in the 1980s, it distinguished limited monitoring from “twenty-four hour surveillance of any citizen in the country,” reserving the question of whether such “dragnet type law enforcement practices” merit the application of different constitutional principles. footnote10_j4487m8 10 See Knotts, 460 U.S. at 283–84.

More recently, the Court’s application of the Fourth Amendment has evolved significantly in response to technological “innovations in surveillance tools.” footnote11_z65ic9i 11 Carpenter, 138 S. Ct. at 2214. In Kyllo v. U.S. (2001), for instance, the Supreme Court held that police need a warrant before they can use a thermal imager to detect heat coming from a garage. By doing so, the Court rejected a return to a “mechanical interpretation” of the Fourth Amendment, under which the Constitution would have protected only against physical intrusions into a person’s private space, holding instead that it was necessary to ensure that people were not left “at the mercy of advancing technology.” footnote12_ex6gj6m 12 Kyllo v. United States, 533 U.S. 27, 35 (2001). Over time, the Court has ruled that law enforcement must obtain a warrant before searching a suspect’s cell phone during an arrest (even though it had previously allowed warrantless searches incident to arrest), before installing a GPS tracker on an automobile for long-term monitoring (despite precedent suggesting that vehicular movements are not private), and before obtaining historical cell-site location information revealing an individual’s daily movements (although third-party information can normally be obtained without a warrant). footnote13_7zc8tcg 13 Riley v. California, 134 S. Ct. 2473 (2014); United States v. Jones, 132 S. Ct. 945 (2012); Carpenter, 138 S. Ct. 2206.

The reasoning in these cases is instructive. Take U.S. v. Jones (2012), in which the Supreme Court held that the police need a warrant in order to install a GPS tracking device on a car and use it for extended surveillance. In her concurrence, Justice Sonia Sotomayor observed that inexpensive location tracking “makes available at a relatively low cost such a substantial quantum of intimate information about any person whom the Government, in its unfettered discretion, chooses to track” that it “may ‘alter the relationship between citizen and government in a way that is inimical to democratic society.’” footnote14_80cu8jk 14 Jones, 132 S. Ct. at 956 (Sotomayor, J., concurring) (quoting United States v. Cuevas-Perez, 640 F.3d 272, 285 (7th Cir. 2011) (Flaum, J., concurring)). Similar themes run through the Court’s decision in Carpenter v. U.S. (2018), which holds that police must get a warrant before they can obtain historical information from cell phone providers about the location of individuals’ mobile phones (known as cell-site location information, or CSLI). footnote15_6xx23cg 15 Carpenter, 138 S. Ct. at 2223. The Court observed that this information could be used to track the minutiae of people’s daily lives. It reasoned that the “depth, breadth, and comprehensive reach” of this data, along with “the inescapable and automatic nature of its collection” by virtue of simply carrying a cell phone, necessitate a warrant supported by probable cause. footnote16_8y27woe 16 Carpenter, 138 S. Ct. at 2223.

While the Carpenter decision narrowly addresses the use of historical CSLI, it provides an important framework for analyzing reasonable expectations of privacy in the digital age. Location tracking via ALPR databases raises many of the same concerns outlined in Carpenter; an application of its framework should lead courts to conclude that police must first obtain a warrant before searching historical location information from ALPR databases.

Specifically, first, Carpenter instructs courts to consider the capacity of a technology to enable ongoing surveillance that would have been unimaginable before the digital age. footnote17_c4ew4f3 17 Carpenter, 138 S. Ct. at 2218. Just as with CSLI, automatic license plate readers enable data collection that is “detailed, encyclopedic, and effortlessly compiled.” footnote18_3yx4td9 18 Carpenter, 138 S. Ct. at 2216. A person’s phone is constantly creating records simply by being powered on and connecting to the network. Similarly, ALPRs automatically collect information about every car that passes within their range. But while a person might turn off their cell phone while they travel, it may be almost impossible to avoid traveling some roads without exposing one’s vehicle to ALPRs.

Second, the Carpenter Court considered the extent to which data collection is indiscriminate, targeting not only people under investigation but a much broader segment of the population. footnote19_wlth70z 19 Carpenter, 138 S. Ct. at 2216. While ALPR scans provide a different level of pinpoint accuracy than CSLI, they also indiscriminately collect data about every car that passes by a license plate reader, regardless of the driver’s connection to criminal activity. In fact, the vast majority of scans capture information about drivers who are not suspected of any wrongdoing. footnote20_zlmxwcd 20 See, e.g., Auditor of the State of California, Automated License Plate Readers, 1 (finding that “99.9 percent of the 320 million images Los Angeles stores are for vehicles that were not on a hot list when the image was made”). The only limitations on this ongoing surveillance of all cars traveling a public road are the number of ALPRs and the data retention policies maintained by police or third-party vendors.

Third, the Court considered the extent to which the long-term CSLI retention allowed officers to effectively create a time machine of a person’s movements. Just as with historical CSLI, the long-term retention of plate data allows the police to retroactively track every location where a particular car was tagged by an ALPR device. footnote21_o51i7e0 21 Carpenter, 138 S. Ct. at 2210. To be sure, the current scope of ALPR devices does not match the scope of cell phone towers blanketing the country, which makes a direct comparison difficult. Nonetheless, the current adoption rate of ALPRs suggests that this technology will continue to expand its coverage areas. In fact, the Carpenter Court ruled that lower courts “must take account of more sophisticated systems that are already in use or in development.” footnote22_ipcgzj4 22 Carpenter, 138 S. Ct. at 2218–19 (quoting Kyllo, 533 U.S. at 36). ALPR technology is expanding at a rapid rate, with growing databases containing billions of license plate scans, and with governmental and private ALPR devices capturing larger swaths of cities. Courts should consider this foreseeable future when confronted with nascent uses of ALPR that appear smaller in scale.

Finally, the Court ruled that an interpretation of the Fourth Amendment called the third-party doctrine is inapplicable to historical CSLI. footnote23_4nxtlqi 23 Carpenter, 138 S. Ct. at 2219–20. Under the third-party doctrine, individuals do not have a reasonable expectation of privacy in information they are deemed to have voluntarily handed over to third parties. footnote24_ytiz1dj 24 See United States v. Miller, 425 U.S. 435, 443 (1976) (“The Fourth Amendment does not prohibit the obtaining of information revealed to a third party and conveyed by him to Government authorities, even if the information is revealed on the assumption that it will be used only for a limited purpose and the confidence placed in the third party will not be betrayed”). The Carpenter Court found that while this doctrine is appropriate for limited disclosures such as bank records or a log of dialed telephone numbers, it should not apply to CSLI data, which can provide a “chronicle of a person’s physical presence compiled every day, every moment, over several years.” footnote25_amcjuh5 25 Carpenter, 138 S. Ct. at 2220. Historical ALPR data similarly chronicles the movements of all vehicles, regardless of the registered owner’s connection to a suspected crime.

The Carpenter Court also reasoned that individuals do not truly voluntarily share their location data with wireless carriers; instead, the data is automatically collected simply by possessing a cell phone — a device the Court described as “indispensable to participation in modern society” — and by connecting to a mobile network. footnote26_ab7innd 26 Carpenter, 138 S. Ct. at 2210. Similarly, a majority of Americans rely on driving in order to fully participate in society, and their movements are logged by ALPRs by virtue of simply driving and parking on public roads. Just as the only way to avoid generating CSLI would be to turn off a mobile device, the only way to avoid ALPR data collection would be to give up driving altogether or to keep a vehicle away from the range of a license plate reader — an impossible task in many places. footnote27_be1g4nj 27 See Department of Homeland Security, “Privacy Impact Assessment for CBP License Plate Reader Technology,” Washington, DC, July 6, 2020, 8, https://www.dhs.gov/publication/dhscbppia-049-cbp-license-plate-reader-technology (“The only way to opt out of such surveillance is to avoid the impacted area, which may pose significant hardships and be generally unrealistic”). Carpenter thus suggests that the third-party doctrine is equally inapplicable to historical location data collected by ALPR devices.

Although the Supreme Court has not yet addressed whether police access to historical ALPR data requires a warrant, appeals courts have begun hearing challenges to warrantless ALPR database searches. However, courts appear reluctant to embrace a bright-line rule that extends Carpenter to ALPR searches. The result has been a series of one-off decisions that seek to avoid direct engagement with the foreseeable proliferation of ALPR data.

For example, in U.S. v. Yang (2020), the Ninth Circuit ruled that the defendant did not have standing to challenge government queries of a private ALPR database for records of his rental car travels when he kept the vehicle past the contract due date in violation of company policy. footnote28_xj4x1cf 28 United States v. Yang, No. 18–10341, 2020 WL 2110973, at *6–7 (9th Cir. May 4, 2020), https://cdn.ca9.uscourts.gov/datastore/opinions/2020/05/04/18–10341.pdf. This ruling now compels defendants in the Ninth Circuit to prove that they had a sufficiently close relationship with the property that was searched before the court will address their Fourth Amendment rights.

And in Commonwealth v. McCarthy (2020), the Massachusetts Supreme Judicial Court ruled that while widespread use of ALPR devices can implicate a person’s Fourth Amendment privacy interest in the whole of their movements, the limited surveillance undertaken in that case did not violate the defendant’s reasonable expectation of privacy. footnote29_l3eujby 29 Commonwealth v. McCarthy, 484 Mass. 493, 494 (2020), https://cases.justia.com/massachusetts/supreme-court/2020-sjc-12750.pdf?ts=1587124946. The case involved police officers’ use of ALPR hot list notifications to track the defendant’s movements as he traveled across two bridges over the course of two months. footnote30_arl8kwy 30 McCarthy, 484 Mass. at 495. The court applied what is commonly referred to as the “mosaic theory,” where the long-term surveillance of a person’s public movements triggers a privacy interest that could be absent with only limited or isolated monitoring. footnote31_uz7iat3 31 McCarthy, 484 Mass. at 502. The court acknowledged that “with enough cameras in enough locations, the historic location data from an ALPR system in Massachusetts would invade a reasonable expectation of privacy and constitute a search for constitutional purposes.” footnote32_egnigl1 32 McCarthy, 484 Mass. at 506. Four ALPR devices on two bridges did not, however, rise to this level, according to the ruling. footnote33_6pum7ru 33 McCarthy, 484 Mass. at 508–9. Lower courts will continue to face the dilemma of how to rule on the specific use of ALPRs in a given case while taking into account the Supreme Court’s admonition that courts must consider the logical evolution of these systems of surveillance. footnote34_50uy9o9 34 See Carpenter, 138 S. Ct. at 2210 (“At any rate, the rule the Court adopts ‘must take account of more sophisticated systems that are already in use or in development,’ . . . and the accuracy of CSLI is rapidly approaching GPS-level precision.” (quoting Kyllo, 533 U.S. at 36)).

Separate from Fourth Amendment considerations, courts have also considered how ALPR technology may violate privacy protections under state law. For example, the Virginia Supreme Court is currently hearing an appeal seeking to reopen the substantive issue of whether the Fairfax County Police Department’s use of an ALPR system to passively track the movements of cars that were not on a hot list violates the state’s Government Data Collection and Dissemination Practices Act. footnote35_9exywme 35 See Petition for Appeal, Neal v. Fairfax County Police Department, Case No. CL-2015–5908, 295 Va. 334 (August 29, 2019), https://rrbmdk.egnyte.com/dl/gCZYOWOzfH. This act requires, among other things, that information not be collected unless the need for it has been clearly established ahead of its collection — a standard that indiscriminate collection of ALPR data cannot meet. footnote36_mdpls5f 36 Government Data Collection and Dissemination Practices Act, VA Code Ann. § 2.2–3800(C)(7). If the trial court’s ruling is upheld, the Fairfax County Police will be required to purge ALPR data that is not linked to a criminal investigation and to stop using ALPRs to passively collect data on people who are not suspected of criminal activity. footnote37_a0n0uuy 37 Petition for Appeal, Neal v. Fairfax County Police Department, 4.

ALPRs are relevant to more than privacy. Courts have also considered whether an ALPR hit provides sufficient justification for a police officer to stop a car. In Kansas v. Glover (2020), the Supreme Court ruled that a license plate search indicating that a car’s registered owner has had his or her license revoked gives police reasonable suspicion to perform a traffic stop in the absence of information suggesting that someone other than the owner is driving the vehicle. footnote38_88asbg3 38 Kansas v. Glover, 140 S. Ct. 1183 (2020). Several state courts reached the same conclusion. footnote39_dmy0jdo 39 See, e.g., Hernandez-Lopez v. State, 319 Ga. App. 662 (2013) (holding that an ALPR alert indicating that a car’s registered owner had an outstanding warrant gave officers reasonable suspicion to justify a traffic stop); see also Hill v. State, 321 Ga. App. 817 (2013) (holding that an ALPR alert from a device mounted on a police cruiser gave an officer reasonable suspicion to justify a traffic stop); and Traft v. Commonwealth, 539 S.W.3d 647 (Ky. 2018) (holding that an ALPR alert indicating an active bench warrant gave an officer reasonable suspicion to justify a traffic stop).

An ALPR hit is not always a sufficient basis for a stop, however. footnote40_qalu01y 40 Green v. City and County of San Francisco, 751 F.3d 1039, 1045 (9th Cir. 2014) (ruling that “an unconfirmed hit on the ALPR does not, alone, form the reasonable suspicion necessary to support an investigatory detention”). In 2014, the U.S. Court of Appeals for the Ninth Circuit considered an erroneous ALPR alert that led to a traffic stop in which a woman was detained and held at gunpoint. footnote41_6zquql5 41 Green, 751 F.3d 1039 at 1041. Unlike in Glover, where an officer manually searched a license plate number and confirmed that the truck he observed matched the vehicle in the database, in Green v. City and County of San Francisco (2014), an ALPR device mounted on an officer’s cruiser malfunctioned and returned a hit for a different vehicle and license plate number than the plaintiff’s car. footnote42_j8fz18r 42 Green, 751 F.3d 1039 at 1046. An officer radioed in a description of the plaintiff’s vehicle, along with the incorrect license plate number picked up by the ALPR device. footnote43_nx41trh 43 Green, 751 F.3d 1039 at 1042. A second officer identified the plaintiff’s car, but did not attempt to confirm whether the radioed license plate number matched the plates on the plaintiff’s car. The Ninth Circuit ruled that the case could proceed on the question of whether the second officer should have taken additional steps to independently confirm whether the ALPR device had identified the right car and license plate number before initiating a traffic stop. footnote44_m34k287 44 Green, 751 F.3d 1039 at 1043. The Green decision, which analyzed an ALPR system that “frequently” makes mistakes, may suggest that there are situations in which reliance on an ALPR hit remains insufficient to justify a traffic stop.

In light of the wide saturation of license plate readers, it is critical that the use of these devices be accurate, bias-free, and protective of established legal values and constitutional rights. Unfortunately, publicly available information suggests that this is not the case. This may explain why at least 16 states have passed laws regulating the use of ALPRs or the use of data collected by the devices. footnote1_j8jms1l 1 “Automated License Plate Readers: State Statutes,” National Conference of State Legislatures, Washington, DC, updated June 23, 2020, https://www.ncsl.org/research/telecommunications-and-information-technology/state-statutes-regulating-the-use-of-automated-license-plate-readers-alpr-or-alpr-data.aspx. Some prohibit the use of ALPRs except for limited public safety purposes, whereas others establish controls governing their use, including mandatory privacy policies, limits on data retention, express limits on the types of investigations in which they can be used, and mandatory audits. footnote2_ye2jlad 2 See, e.g., Ark. Code Ann. §§ 12–12–1803 (prohibiting use of ALPRs by individuals, partnerships, corporations, associations, or state agencies, and limits use by law enforcement); Neb. Rev. Stat. §§ 60–3206 (requiring any state agency using ALPRs to adopt and post a privacy policy); Mont. Code Ann. §§ 46–5–118 (establishing a ninety-day limit on the retention of license plate data collected by ALPRs); 23 Vt. Stat. Ann. §§ 1607–07 (“Deployment of ALPR equipment by Vermont law enforcement agencies is intended to provide access to law enforcement reports of wanted or stolen vehicles and wanted persons and to further other legitimate law enforcement purposes.”); and Md. Public Safety Code § 3–509(b)(1)(2)(c)(2)(ii) (establishing mandatory audits of ALPR systems). These regulations highlight many of the concerns around ALPRs listed below and predict many of the recommendations that follow.

• High error rates: Errors can arise in at least two ways — inaccurate hot lists and inaccurate reads. If hot lists are not updated, an individual may be pulled over when, for instance, the system incorrectly indicates that a license is suspended when it has actually been reinstated. Inaccurate reads are surprisingly common as well: one randomized control trial in Vallejo, California, found that 37 percent of all ALPR “hits” from fixed readers (such as those attached to a street light) and 35 percent from mobile ALPRs were misreads — an astonishingly high error rate. footnote3_rhxzdiq 3 See generally, Davila, 901 N.Y.S.2d 787; see also Jason Potts, “Research in Brief: Assessing the Effectiveness of Automatic License Plate Readers,” Police Chief March 2018, 14, https://www.theiacp.org/sites/default/files/2018–08/March%202018%20RIB.pdf. In several high-profile incidents, drivers have been pulled over because a reader read the numbers on their license plates wrong and erroneously tagged the vehicles as stolen. footnote4_g5lel2c 4 Tim Cushing, “Deputies Sued after False ALPR Hit Leads to Guns-Out Traffic Stop of California Privacy Activist,” TechDirt (Floor64 blog), February 20, 2019, https://www.techdirt.com/articles/20190217/08240241618/deputies-sued-after-false-alpr-hit-leads-to-guns-out-traffic-stop-california-privacy-activist.shtml; and Cyrus Farivar, “Due to License Plate Reader Error, Cop Approaches Innocent Man, Weapon in Hand” Ars Technica, April 23, 2014, http://arstechnica.com/tech-policy/2014/04/due-to-license-plate-reader-error-cop-approaches-innocent-man-weapon-in-hand. In one instance, a Colorado woman and several children were detained and handcuffed facedown on the ground after an ALPR mistook their SUV for a stolen motorcycle from a different state. footnote5_gfjd70t 5 Jessica Porter, “Aurora Police Detain Black Family after Mistaking Their Vehicle as Stolen,” ABC7 Denver, August 3, 2020, https://www.thedenverchannel.com/news/local-news/aurora-police-detain-black-family-after-mistaking-their-vehicle-as-stolen. In another, the chair of Oakland’s Privacy Advisory Commission was mistakenly stopped and detained at gunpoint after his rental car’s license plate triggered an out-of-date hit signaling to police officers that the car had been stolen. footnote6_4pqyf2r 6 See Lisa Fernandez, “Privacy Advocate Sues CoCo Sheriff’s Deputies after License Plate Reader Targets His Car Stolen,” KTVU Fox 2, February 19, 2019, https://www.ktvu.com/news/privacy-advocate-sues-coco-sheriffs-deputies-after-license-plate-readers-target-his-car-stolen. Even in cases where a vehicle is accurately flagged, it may not convey accurate information about an individual. A car can be shared among family members, among friends, or as part of a carshare; this reality may place low-income individuals, who are more likely to share cars, at greater risk of misidentification.

To be sure, license plate readers have had some high-profile successes: a man accused of stabbing several people after breaking into a rabbi’s home during a Hanukkah celebration was found in part due to an alert from an ALPR device; a Tennessee girl abducted by her noncustodial father was recovered when a license plate camera spotted his car; and police were able to use information from a license plate reader to help halt a string of random shootings on highways in Kansas City, Missouri. footnote7_ahwkg9r 7 Jonathan Dienst et al., “Suspect in NY Hanukkah Stabbings Researched Hitler, Had 2 Knives in Car: Official,” NBC4 New York, December 30, 2019, https://www.nbcnewyork.com/news/local/what-we-know-about-monsey-stabbing-suspect/2252284/; Carly Moore, “License Plate Readers Used in Finding Missing Tennessee Girl,” KKCO NBC11 News, August 19, 2016, http://www.nbc11news.com/content/news/390756761.html; and Matt Pierce, “How Technology Helped Crack the Kansas City Highway Shooter Case,” Los Angeles Times, April 20, 2014, http://www.latimes.com/nation/nationnow/la-na-nn-kansas-city-highway-shooter-20140419-story.html. Despite these anecdotal successes, there has not been a thorough assessment of the tool’s value. Any such assessment would require consideration of the ALPR’s additional costs and benefits described here.

• Privacy and data security concerns: An extremely small percentage of cars scanned by ALPRs — generally far below 1 percent — are connected to any crime or wrongdoing. footnote8_m289yct 8 See, e.g., Maass and Lipton, “Data Driven: Explore How Cops Are Collecting and Sharing Our Travel Patterns”; see also Crump, You Are Being Tracked, 13–15 (noting that “only 0.2 percent of reads [in Maryland] were associated with any crime, wrongdoing, minor registration problem, or even suspicion of a problem,” and 0.05 percent of plate reads by the Minnesota State Patrol led to citations or arrests). For example, an audit found that 99.9 percent of the ALPR images stored by the LAPD are for vehicles not on a hot list at the time a license plate was scanned. footnote9_9cy48wd 9 Auditor of the State of California, Automated License Plate Readers, 1. Nevertheless, many jurisdictions keep the scans “just in case,” storing the data for anywhere from 90 days to two years or even indefinitely. footnote10_a4wgks6 10 See, e.g., Crump, You Are Being Tracked, 8–9 (noting that a 2012 survey of 40 police departments found that 5 percent did not retain ALPR reads; 18 percent retained data for either 30 days or less or for between two and six months; and 13 percent retained ALPR reads indefinitely); Seattle Police Department, 2018 Surveillance Impact Report: Automated License Plate Recognition (ALPR) (Patrol), Seattle Information Technology, Seattle, WA, January 31, 2019, 21, http://www.seattle.gov/Documents/Departments/Tech/Privacy/SPD%20ALPR%20(Patrol)%20-%20Final%20SIR.pdf (noting that data is retained for 90 days); San Francisco Police Department, “Automated License Place Recognition Vehicles,” Department Bulletin, San Francisco, CA, September 22, 2010, https://cdn.muckrock.com/foia_files/2019/02/08/ALPR20DB20DGO20POLICIES.pdf (noting that plate scan information is retained for two years); Los Angeles County Sheriff’s Department, “Automated License Plate Recognition (ALPR) Privacy Policy,” Los Angeles, CA, accessed August 26, 2020, https://web.archive.org/web/20190711152351/http://shq.lasdnews.net/content/uoa/EPC/ALPRPrivacyPolicy.pdf (noting that plate scan information is retained for two years); and Shawn Musgrave, “Under Current Policy, Boston Police Can Keep Scanned License Plate Data Indefinitely,” Muckrock, October 1, 2012, https://www.muckrock.com/news/archives/2012/oct/01/under-current-policy-boston-police-can-keep-scanne. These scans, over time, can reveal individuals’ movements and help create detailed pictures of their private lives. footnote11_r9x734r 11 See Department of Homeland Security, “Privacy Impact Assessment for CBP License Plate Reader Technology,” 10 (“[A]LPR data from third party sources may, in the aggregate, reveal information about an individual’s travel over time, or provide details about an individual’s private life, leading to privacy concerns or implicating constitutionally-protected freedoms.”), https://www.dhs.gov/sites/default/files/publications/privacy-pia-cbp049a-cbplprtechnology-july2020.pdf.

In addition to information generated by ALPRs, police officers can also add to and store sensitive information in the databases housing license plate scans through open text fields and hot lists available in the user interface. For example, the California state auditor found that law enforcement can input information including personal information such as names, addresses, dates of birth, and physical descriptions, and they can also store criminal justice information such as criminal charges and warrant information. footnote12_jlctr8l 12 Auditor of the State of California, Automated License Plate Readers, 18. License plate readers have also been known to capture private information, such as shots of children exiting a car in the driveway of a home or activity inside an open garage — information that surely should not be retained. footnote13_0m9mc5u 13 See Ali Winston, “License-Plate Readers Let Police Collect Millions of Records on Drivers,” Reveal News (Center for Investigative Reporting, Berkeley, CA), June 26, 2013, https://www.revealnews.org/article-legacy/license-plate-readers-let-police-collect-millions-of-records-on-drivers. This is information that goes far beyond the legitimate need to find stolen cars or vehicles linked to AMBER Alerts. The ongoing storage of this wide array of sensitive information also raises security concerns, as this information can be vulnerable to data breaches and hacking. The data security applied to ALPR data may not be commensurate with the sensitivity of the data being held.

• Data sharing concerns: Many vendors allow their law enforcement clients to share and receive ALPR data from other law enforcement agencies. For example, through Vigilant Solutions’ Law Enforcement Archival Reporting Network (LEARN), police departments can elect to automatically share their collection of license plate reads with outside law enforcement partners that are also part of the network. These data sharing arrangements are not always made public or adequately tracked by police departments, which can result in impermissible or unaccountable sharing. An ACLU investigation found that more than 80 local police departments had set up their LEARN settings to share ALPR data with U.S. Immigrations and Customs Enforcement (ICE), even though the practice may violate local privacy laws or sanctuary policies. footnote14_4ka8sad 14 See Neb. Rev. Stat. § 60–3206. Local laws and policies will have limited effect if they do not address automated data sharing or if law enforcement cannot effectively control data flows in and out of their departments. In one instance, the California state auditor found that despite efforts to limit data sharing with ICE, confusing vendor settings had left three different ICE agencies with access to ALPR data from Marin County Sheriff’s Office, frustrating compliance with a California law that places controls on local police cooperation with immigration authorities. footnote15_gcxkzfo 15 Auditor of the State of California, Automated License Plate Readers, 26. Customs and Border Protection also receives ALPR data from commercial vendors, including information from across the United States, “outside of the border zone in which CBP activities take place.” footnote16_wwlpfrl 16 Department of Homeland Security, “Privacy Impact Assessment for CBP License Plate Reader Technology,” 6.

With public agencies seeking to collect ALPR data for uses such as toll collection or environmental analysis, there are concerns that the information being collected may be intentionally or unintentionally shared with law enforcement agencies. Without policies governing the type of data that is collected, stored, and shared, these government data sets may create a frictionless data sharing opportunity that frustrates attempts to limit law enforcement access to ALPR data. For example, in San Diego, law enforcement officers regularly access smart streetlight footage — in some cases, to surveil Black Lives Matter protests — even though the streetlights project was originally intended to assist city planners and app developers. footnote17_ttil3xd 17 Jesse Marx, “Police Used Smart Streetlight Footage to Investigate Protesters,” Voice of San Diego, June 29, 2020, https://www.voiceofsandiego.org/topics/government/police-used-smart-streetlight-footage-to-investigate-protesters/.

As more companies sell ALPRs to homeowners, additional data sharing concerns emerge. For example, this trend allows police officers to expand the reach of their surveillance systems by providing them with access to private device feeds that may be outside the scope of law enforcement policies governing their own equipment (if any exist at all). When police officers solicit data from private ALPR systems, the decision to share information is up to the individual homeowner or the private company providing the service. While companies may maintain privacy policies that explain the situations in which they share information with law enforcement, the policy only covers the company and the person who purchased the devices. footnote18_11bz4gd 18 See, e.g., “Privacy Policy for Flock Safety,” Flock Safety Group, Inc. (website), Atlanta, GA, updated February 22, 2017, https://www.flocksafety.com/legal/privacy-policy (“We may also access, use, preserve and/or disclose your personal information or Recordings to law enforcement authorities, government officials, and/or third parties, if legally required to do so or if we have a good faith belief that such access, use, preservation or disclosure is reasonably necessary to: (a) comply with a legal process or request; (b) enforce our Terms of Service, including investigation of any potential violation thereof; (c) detect, prevent or otherwise address security, fraud or technical issues; or (d) protect the rights, property or safety of Flock, its users, a third party, or the public as required or permitted by law.”). The registered owners of vehicles tracked and logged by these private devices will not receive notice or an opportunity to object to data sharing arrangements between police and private individuals.

• Layering ALPR data with other surveillance systems: License plate readers can be used alongside other kinds of technologies to facilitate even more widespread surveillance. The New York City Police Department (NYPD) Domain Awareness System, for instance, can track the movements of cars and people using its 20,000 security cameras and license plate readers (along with face and object detection technology). footnote19_26aujjh 19 Rocco Parascandola and Tina Moore, “NYPD Unveils New $40 Million Super Computer System that Uses Data from Network of Cameras, License Plate Readers and Crime Reports,” NY Daily News, August 8, 2012, https://www.nydailynews.com/new-york/nypd-unveils-new-40-million-super-computer-system-data-network-cameras-license-plate-readers-crime-reports-article-1.1132135; see also “A Conversation with Jessica Tisch ’08: Revolutionizing Police Tech with the NYPD,” Harvard Law Today, July 17, 2019, https://today.law.harvard.edu/a-conversation-with-jessica-tisch-08 (claiming that the NYPD has access to 20,000 CCTV cameras). Several police departments, including those in Chicago, Detroit, and Memphis, incorporate ALPR technology into Real Time Crime Centers (RTCCs) that combine license plate data with surveillance footage from thousands of police, school, and traffic cameras, gunshot detection systems, and social media monitoring. footnote20_kfbhm3a 20 Carol Robinson, “Birmingham Police Will Fight Crime with All-Seeing Live Technology,” AL.com, May 28, 2019, https://www.al.com/news/birmingham/2019/05/birmingham-police-will-fight-crime-with-all-seeing-live-technology.html. The expansive and ongoing monitoring that is facilitated through these integrated systems may be incompatible with constitutional freedoms, including the right to free assembly and the right to privacy. These burdens hinder the collective organizing necessary to hold the government accountable to the will of the people.

• Lack of transparency and access controls: While ALPRs are increasingly ubiquitous, many police departments do not actively maintain use policies, and there are insufficient controls to protect against misuse. footnote21_c1qo25i 21 See, e.g., Shawn Musgrave, “Massachusetts Police Lack Policies for License Plate Scanners,” Muckrock, April 10, 2013, https://www.muckrock.com/news/archives/2013/apr/10/license-plate-scanners-use-across-massachusetts/. A 2016 investigation by the Associated Press found that police officers across the country abuse confidential databases to spy on love interests, journalists, business associates, and others. footnote22_q87b3jm 22 Sadie Gurman, “AP: Across US, Police Officers Abuse Confidential Databases,” Associated Press, September 27, 2016, https://apnews.com/699236946e3140659fff8a2362e16f43/ap-across-us-police-officers-abuse-confidential-databases. For example, the investigation reported officers stalking ex-girlfriends, looking up the addresses of crushes, and in one case, running searches on a journalist who wrote a series of stories critical of the department. ALPR databases could easily be put to similar use. footnote23_1ame9uj 23 Gurman, “Police Officers Abuse Confidential Databases.” Police departments exacerbate this problem when they fail to implement access and monitoring safeguards to ensure that ALPR data is only accessed on a “need to know” basis, and that access is appropriately logged and monitored to protect against misuse. Instead, some departments automatically install ALPR software on every computer assigned to staff, even when their position does not require access to this kind of information. footnote24_n8cw3be 24 See, e.g., Auditor of the State of California, Automated License Plate Readers, 33. There is a further lack of transparency in many jurisdictions where vendor contracts prohibit police departments from disclosing their use of surveillance systems to the public. footnote25_mn7bxao 25 See, e.g., Cyrus Farivar, “FBI Really Doesn’t Want Anyone to Know About 'Stingray’ Use by Local Cops,” ArsTechnica, February 10, 2015, https://arstechnica.com/tech-policy/2015/02/fbi-really-doesnt-want-anyone-to-know-about-stingray-use-by-local-cops/.

• Disparate impact concerns: ALPRs can also be deployed to target communities of color or other vulnerable populations. The NYPD has used license plate readers as part of its widespread surveillance of Muslim communities in the New York and New Jersey area. footnote26_uzirj9k 26 Adam Goldman and Matt Apuzzo, “NYPD Defends Tactics over Mosque Spying; Records Reveal New Details on Muslim Surveillance,” Huffington Post, February 25, 2012, http://www.huffingtonpost.com/2012/02/24/nypd-defends-tactics-over_n_1298997.html; see also Hassan v. City of New York, 804 F.3d 277, 285–87 (3d Cir. 2015) (listing factual allegations by plaintiffs alleging extensive, targeted surveillance of Muslim community); Diala Shamas and Nermeen Arastu, Mapping Muslims: NYPD Spying and Its Impact on American Muslims, Muslim American Civil Liberties Coalition, Creating Law Enforcement Accountability and Responsibility, and Asian American Legal Defense and Education Fund, New York, NY, 2013, 14, https://www.law.cuny.edu/wp-content/uploads/page-assets/academics/clinics/immigration/clear/Mapping-Muslims.pdf. And an investigation of license plate readers in Oakland, California, found that they were located predominantly in Black and Latino neighborhoods, despite the fact that automobile crimes and offenses predominantly occurred elsewhere. footnote27_sbc8il7 27 Dave Maass and Jeremy Gillula, What You Can Learn from Oakland’s Raw ALPR Data, Electronic Frontier Foundation, San Francisco, CA, January 21, 2015, https://www.eff.org/deeplinks/2015/01/what-we-learned-oakland-raw-alpr-data. Even the placement of ALPRs in “high crime” neighborhoods will likely reflect a history of biased and selective enforcement that has already led to the over-policing of communities of color. The guise of neutral surveillance will only reinforce these practices and maintain the attendant potential for deadly police encounters.

Some police departments also incorporate ALPR data into gang databases, which allows officers to track vehicles associated with suspected gang members. footnote28_joii2fr 28 See, e.g., City of Chicago Office of Inspector General, Review of the Chicago Police Department’s “Gang Database,” Chicago, IL, April 2019, https://igchicago.org/wp-content/uploads/2019/04/OIG-CPD-Gang-Database-Review.pdf. These gang databases are notoriously unreliable, as they rely on vague and often contradictory criteria for inclusion. footnote29_2493m26 29 See, e.g., Anita Chabria, “A Routine Police Stop Landed Him on California’s Gang Database. Is It Racial Profiling?,” Los Angeles Times, May 9, 2019, https://www.latimes.com/politics/la-pol-ca-california-gang-database-calgang-criminal-justice-reform-20190509-story.html. Gang databases, which contain tens of thousands of names, are almost overwhelmingly comprised of individuals of color, and people frequently have no opportunity to challenge their inclusion. footnote30_gx81p48 30 See Alice Speri, “New York Gang Database Expanded by 70 Percent Under Mayor Bill de Blasio,” Intercept, June 11, 2018, https://theintercept.com/2018/06/11/new-york-gang-database-expanded-by-70-percent-under-mayor-bill-de-blasio/. The LAPD suspended use of California’s statewide gang database after announcing audits and investigations in response to allegations that police falsified records and listed innocent people as gang members. footnote31_uko31rw 31 Kristina Bravo, “LAPD Suspends Use of CalGang Database Months after Announcing Probe of Officers Accused of Falsifying Information,” KTLA Los Angeles, January 20, 2020, https://ktla.com/news/local-news/lapd-suspends-use-of-calgang-database-months-after-announcing-probe-of-officers-accused-of-falsifying-information/. Meanwhile, an audit by Chicago’s Office of Inspector General found that the city’s gang database contained incomplete and conflicting data, with some entries raising serious concerns about how officers “perceive and treat the people with whom they interact.” footnote32_kjwfc0j 32 City of Chicago Office of Inspector General, Review of the Chicago Police Department’s “Gang Database,” 2, 6.

In the wake of nationwide protests that followed the police killings of George Floyd and Breonna Taylor, public attention has increasingly focused on the ongoing instances of police brutality and racial bias in policing. However, there is a risk that police departments and legislators may incorrectly propose surveillance as a neutral alternative. Surveillance that disproportionately targets communities of color carries a distinct and cognizable equal protection harm: branding them with a badge of inferiority. As one appellate court wrote, “Our nation’s history teaches the uncomfortable lesson that those not on discrimination’s receiving end can all too easily gloss over the ‘badge of inferiority’ inflicted by unequal treatment itself. Closing our eyes to the real and ascertainable harms of discrimination inevitably leads to morning-after regret.” footnote33_o8i4frc 33 Hassan, 804 F.3d at 291.

• Impact on protected First Amendment rights: Law enforcement agencies have a history of misusing license plate surveillance to monitor First Amendment–protected activity. During the 2008 presidential election, the Virginia State Police recorded the license plate numbers of attendees at political rallies for Barack Obama and Sarah Palin — and subsequently at President Obama’s inauguration — and kept the data for more than three years until it was purged following an opinion from the Virginia Attorney General warning that ongoing retention would violate the state’s Government Data Collection and Dissemination Practices Act. footnote34_j0zhp9c 34 Mark Bowes, “Police Recorded License Plates at Obama Inauguration,” Richmond Times-Dispatch, August 18, 2013, http://www.timesdispatch.com/news/local/crime/article_32678a59-f9e1–5e46–8336-d5f4ba076cb7.html. Similarly, police in Denver spied on anti-logging activists and shared license plate information with the FBI’s Joint Terrorism Task Force when the activists held a training on nonviolence. footnote35_5p0ex79 35 Kirsten Atkins, “Statement — Kirsten Atkins, Target of Illegal Spying,” American Civil Liberties Union, New York, NY, https://www.aclu.org/statement-kirsten-atkins-target-illegal-spying.

Such surveillance — whether it involves the locations of multiple cars that appear together in the same place, or of a single car at places like a mosque, synagogue, or rally — has a chilling effect on Americans’ First Amendment rights to freedoms of association, religion, and speech. footnote36_togqtg3 36 See, e.g., United States v. Jones, 565 U.S. 400, at 416 (“Awareness that the Government may be watching chills associational and expressive freedom.”). An investigation into an NYPD program that monitored mosque visitors’ license plates found that this surveillance “chilled constitutionally protected rights — curtailing religious practice, censoring speech and stunting political organizing.” footnote37_73h103l 37 Shamas and Arastu, Mapping Muslims, 4. The International Association of Chiefs of Police has noted that ALPRs can cause people to “become more cautious in the exercise of their protected rights of expression, protest, association, and political participation because they consider themselves under constant surveillance.” footnote38_5pm94j6 38 International Association of Chiefs of Police, Privacy Impact Assessment Report for the Utilization of License Plate Readers, 13. And there is always the specter of more flagrant abuse, such as putting a political opponent’s license plate on a hot list and using it to keep track of that person’s whereabouts. footnote39_bc104c1 39 See, e.g., Glenn Greenwald, “Government Harassing and Intimidating Bradley Manning Supporters,” Salon, November 9, 2010, http://www.salon.com/2010/11/09/manning_2.

In light of these concerns, the need for a multifaceted response to the proliferation of ALPR use is overdue. This section contains a number of recommendations for policymakers, law enforcement agencies, and technology vendors. These are intended as starting points, as the circumstances and implementation of ALPR use reforms will vary by jurisdiction.

• Adopt retention limits and require warrants for searching historical data: Plates that are scanned and do not match a hot list alert should be promptly discarded. Using ALPRs to record the movements of all vehicles in a municipality goes far beyond the limited information that is revealed by an isolated capture of license plate data. It also goes beyond what an ordinary person could observe on a public road. Alternatively, if plates are retained, the retention period should be as brief as possible — on the scale of days, not months. These limits could be imposed both by departmental policies and by state law. If municipalities elect to permit retention of license plate data to enable historical searches, such searches should require a warrant supported by probable cause absent emergency situations. ALPR databases collect expansive and sensitive accounts of people’s movements regardless of whether they are suspected of criminal activity in a manner that is not available through traditional surveillance. Aside from ceasing to drive altogether, it is exceedingly difficult to avoid this type of surveillance — a condition that is likely to become more pronounced as ALPR technology continues to expand.

• Institute a two-step scanning process: When it comes to officers individually running plates, police departments should tailor their scanning processes so that the first pass through a database will not yield protected personal information, instead revealing only registration information and presence on a hot list. Only if that first inquiry reveals a “basis for further police action” should the officer be permitted to proceed to a second step, which would “allow access to the ‘personal information’ of the registered owner, including name, address, social security number, and if available, criminal record.” footnote1_6ir467k 1 State v. Donis, 157 N.J. 44, 55 (1998). This simple process can help deter police use of these systems for purposes other than law enforcement.

• Require verification of hot list data: Law enforcement agencies should enact policies that require both independent verification of the information yielded from a hot list and real-time updating of hot list data. These steps would help prevent erroneous and potentially dangerous stops based on incorrect or outdated information.

• Require transparency and invite community input into ALPR use and policies: The public should have an opportunity to offer input into whether and how ALPRs are deployed. Given that a very small percentage of license plates scanned will actually be connected to a crime, communities may decide that money spent on ALPRs is better spent on other public safety needs, or that the privacy trade-offs are not worthwhile. If ALPRs are purchased, mechanisms must be in place to solicit feedback from interested community members on the policies governing their use. This public consultation process should also inform the types of crimes that merit inclusion on an ALPR hot list. Draft and final policies should be easily available to the public and must include the results of ongoing audits to detect and deter misuse of the system. ALPRs should not be deployed absent clear and enforceable use policies. Except in emergency circumstances, historical ALPR searches should not be conducted absent a warrant supported by probable cause.

• Maintain audit logs: Law enforcement use of ALPR data should be logged and stored in a format that permits auditing. First, a log should maintain details about automated ALPR alerts, including the reason for the alert, whether any information was automatically shared with other agencies, and the outcome of the alert. Second, logs should track every time an officer seeks to access historical ALPR data as part of an investigation. This log should specify the officer and the crime being investigated and require evidence that a warrant has been obtained or specification of the exigent circumstances that mandate quicker access. If this functionality is not available through vendor platforms, law enforcement should establish internal access controls to ensure the same outcome.

Separately, each police department should establish a log that tracks and catalogs all the ways they receive, store, and share ALPR data. This includes the license plate reads collected by their own devices, as well as those provided by other law enforcement agencies, by private vendors, and voluntarily by businesses and individuals. Many vendor platforms provide automated methods for tracking and updating authorized data flows, but each department should appoint an appropriate office to lead their efforts to track and maintain this log. When a department elects to share ALPR data with another law enforcement agency, the parties should enter into data sharing arrangements ensuring that policies regarding access control and retention are at least as strict as those of the originating agency. The receiving agency should also commit to entering into similar data sharing agreements for any downstream data sharing. Without adequate steps to protect downstream data sharing, even the most rigid policies will be insufficient once data is shared with a department that does not maintain the same level of protection.

• Conduct audits for disparate impact: Law enforcement use of ALPRs should be periodically audited in order to protect against disparate impact on historically marginalized communities and constitutionally protected activities. These audits should evaluate the times and locations where ALPRs are used to ensure that they are not being used to disproportionately target particular communities or constitutionally protected activities such as protests. To facilitate this process, law enforcement agencies must keep records that detail the locations where ALPRs are deployed and the areas where historical searches are being run. Audits should also assess the types of investigations that merit a vehicle’s inclusion on a hot list to ensure that low-level offenses are not effectively being used to target vulnerable communities. Audits should evaluate the extent to which ALPR data is used with other surveillance technologies — such as predictive policing algorithms or inclusion in gang databases — in a manner that could disproportionately harm historically marginalized groups or constitutionally protected activity.
• Conduct audits to ensure effective safeguards: Every ALPR policy should include regular audits to evaluate safeguard effectiveness. These audits should ensure that ALPR data is only available to employees with a need to access the data, that their access is promptly terminated when no longer necessary, and that ALPR searches are appropriately limited to specific law enforcement investigations. Ongoing oversight of the use of ALPR data within law enforcement agencies is an essential safeguard to detect and prevent officers’ misuse of the system.