December 17, 2019
Facial Recognition Regulation – A Year in Review
Clinical Instructional Fellow, Harvard Law School Berkman Klein Center for Internet and Society
Over the course of the last year, lawmakers at the federal, state, and local levels have introduced, and in a few notable cases enacted, a variety of proposals to regulate facial recognition technology. These proposals range from all out bans on government use of facial recognition to narrowly drawn prohibitions on very specific uses of the technology. While any given proposal may leave significant gaps in regulation, the combination of federal, state, and local regulation has the potential to form a robust system of protections. Absent sweeping federal regulation—which seems unlikely in the current political climate—continued action at every level of government is needed to protect our civil liberties.
There are good reasons lawmakers have turned their attention to facial recognition technology. For starters, many facial recognition models show significant bias against marginalized groups. Time and time again, studies have shown that facial recognition tools are less accurate when applied to women and people with darker skin, including in commercial systems used by federal law enforcement agencies. Specific uses of facial recognition raise additional problems—such as automated gender recognition, which regularly misidentifies transgender people. As facial recognition technology proliferates, so do the consequences of misidentification, from being denied boarding for a flight to being wrongfully arrested.
Inaccuracy isn’t the only issue with facial recognition, though. Like many technologies promising improved efficiency, facial recognition can act as a force multiplier for already flawed and unjust systems. This is especially true where facial recognition technology is used to match subjects to existing criminal databases—a common use case for law enforcement agencies. Black and latinx people are more likely to be arrested, and therefore more likely to be included in criminal databases, than white people, even where there are similar levels of criminal conduct. When facial recognition is used match images against these databases, the overinclusion bias towards black and latinx individuals is amplified by the power of modern technology.
Even for those individuals fortunate enough to avoid or opt out of facial recognition systems, there is a very real risk of chilling free speech. The chilling effect of government surveillance is well-documented and is backed up by empirical research. Facial recognition technology, in combination with the proliferation of government-controlled video feeds in the form of surveillance and police body-worn cameras, presents concerns of a constitutional dimension. In addition to chilling free speech in contravention of the First Amendment, government use of facial recognition technology may violate the Fourth Amendment guarantee against unreasonable searches and—due to the issues of bias described above—the Equal Protection clause of the Fourteenth Amendment.
Fortunately, progress is being made. At the local level, 2019 was something of a banner year for the regulation of facial recognition. Since January, San Francisco, Oakland, and two Boston area suburbs have banned municipal use of facial recognition technologies. State governments took up the issue as well. California recently enacted a law placing a three-year moratorium on the use of facial recognition with police-worn body cameras. Michigan is considering a bill that would prohibit law enforcement from using facial recognition in conjunction with any real-time video feed. Proposed legislation in Massachusetts would go even further, placing a moratorium on all government use of facial recognition technology without express statutory permission.
Progress at the state and local level doesn’t obviate the need for federal legislation though. After all, a state-by-state or city-by-city approach could leave significant portions of the population without protection from the harms associated with facial recognition. Moreover, only the federal government can regulate federal agencies such as the FBI and ICE, both of which are known proponents of facial recognition technology. In fact, it was the secretive and potentially unlawful use of facial recognition by these federal agencies that prompted a series of hearings in front of multiple congressional committees earlier this year.
Over the course of the 2019, at least half a dozen bills that would limit the use of facial recognition technology were introduced in Congress. However, none of these bills have yet made it out of committee. Moreover, some of these proposals are extremely limited in scope. Take, for example, Senate bill S.2878. Titled the “Facial Recognition Technology Warrant Act of 2019,” it certainly sounds impressive. It seems to check many of the right boxes: a warrant requirement for the use of facial recognition, suppression of evidence as a remedy, and mandatory reports on the extent of government use of facial recognition. However, its scope is limited to the use of facial recognition for “ongoing surveillance,” defined as the tracking of an individual’s physical location for a period of more than three days.
Civil liberties advocates may find the protections provided by S.2878 underwhelming for a couple of reasons. First, the proscribed use of facial recognition—tracking an individual’s physical movement through public spaces—is not yet widespread. Rather, the most common police use of facial recognition is to match images or video stills against government databases. Second, under the Supreme Court’s reasoning in United States v. Carpenter, tracking using facial recognition likely already requires a warrant. This doesn’t mean S.2878 is not helpful; it could take years for courts to recognize constitutional limitations on facial recognition surveillance, and a narrow bill like S.2878 may be more politically viable than omnibus facial recognition legislation. Still, we should recognize S.2878 for what it is: a narrow fix that, constitutionally speaking, is literally the least we can do.
Another federal proposal that is more immediately applicable is H.4021, which would require federal agencies to get a warrant before running facial recognition matches against government identification databases. An even more sweeping proposal is H.3875, which would prohibit the purchase of facial recognition technology with federal funds. However, H.4021 and H.3875 have not picked up bipartisan support, making their prospects for advancement dim.
Ultimately, it seems clear that advancing the causes of privacy, free speech, and equal protection will require a combination of approaches. Local and state governments have proven more responsive to growing concerns about misuse of facial recognition, but federal action is necessary as well to check the power of the FBI, ICE, and other powerful federal investigative agencies. At every level of government, choices must be made about what uses of facial recognition should be prohibited. While real-time facial analysis and physical tracking are certainly dangers to civil liberties, lawmakers must not lose sight of the more mundane, but also more common, uses of facial recognition as well. A complete ban, or at least a moratorium, on government use of facial recognition may be the safest route for the time being—at least where there is sufficient political will for such regulation. Failing that, a robust network of laws regulating various uses of facial recognition technology at multiple levels of government is essential for the protection of our civil liberties.
Criminal Justice, Electronic Privacy, National Security and Civil Liberties, Race and Criminal Justice, Technology Law and Intellectual Property