Georgetown Law
Technology Review

In March, the Virginia State Legislature approved a bill permitting the use of facial recognition technology (FRT) by local and college campus police departments. If passed, the bill will repeal a temporary ban on such use instituted in 2021.

Although data shows that the public is becoming more accepting of the use of FRT by law enforcement, this bill is too broad because it permits the use of FRT in any investigation if an officer has reasonable suspicion, will result in chilling effects on the expression of First Amendment rights, and will have unequal effects on people of color. While the bill imposes some limitations—it prohibits FRT from being used to create probable cause for an arrest warrant, to compile databases of images derived from video surveillance for future investigations, or to use real-time surveillance—these are not nearly enough.

First, the authorization to use FRT in cases of reasonable suspicion [that] the individual has committed a crime is too broad and ripe for abuse. Virginia should follow the models of other states who have chosen to combat this by taking a more restrictive approach. For example, the Massachusetts Facial Recognition Commission suggests allowing only the Massachusetts State Police to perform FRT searches and limiting its use to felony investigations with a warrant, absent exigent circumstances.  Maine’s bill similarly limits the use of FRT in criminal investigations to investigat[ions of] serious crime when there is probable cause or for fraud prevention or investigation. To balance concerns that courts would be inundated with warrant requests, Virginia should implement a severity requirement for crimes that FRT will be used to investigate. This would limit the subsequent invasion into civilian lives while still furthering significant law enforcement goals.

Second, the breadth of Virginia’s proposed law could inhibit First Amendment expression for those who do not want to be subject to surveillance. Here, too, Virginia deviates from other states’ more lenient bills. While Virginia’s bill makes no mention of such activities, Colorado’s bill prohibits the use of FRT to record a person exercising their First Amendment rights. And Washington’s bill expressly prohibits using FRT to identify individuals based on their participation in lawful activity. Virginia’s bill should protect these important constitutional rights and ensure that persons engaged in these activities cannot be subject to FRT investigation on that basis.

Third, the shortcomings of Virginia’s bill may also have inordinate effects on people of color. In an attempt to confirm accuracy, the bill requires approved FRT to have been assessed by the National Institute of Standards and Technology (NIST) in its “Face Recognition Vendor Test” and have 98 percent or greater true positives in at least one set of data reported in a NIST Face Recognition Vendor Test. The section also requires minimal performance variations across demographics of race, skin tone, ethnicity, or gender. There is debate over how accurate FRT is in identifying women and people of color. Although Virginia’s bill appeals to NIST’s evaluations of FRT, these accuracy scores could be by persons implementing the technology. Also, requiring seemingly high accuracy rates of 98 percent could still cause many misidentifications. For example, in Florida, if 0.1 percent of identifications are incorrect, eight individuals per month may be wrongfully subject to criminal investigation.

Members of Virginia’s legislature have recognized this concern.  Many expressed that FRT could disproportionately impact communities that are already familiar with over-policing and frequent contact with law enforcement. As a response to this, the Virginia bill requires the Department of Criminal Justice Services to review the state’s FRT usage data and prepare a report by November 1, 2025 to include proposals for bettering the use of [FRT] as it relates to race, skin tone, ethnicity, and gender. But because people of color have more interactions with the police, they are more likely to be subject to FRT identification and its shortcomings. Governor Youngkin should require further research into how local agencies can fairly implement the technology across diverse groups and communities before the use is allowed.

Interestingly, few, if any, challenges to the use of FRT have been raised on equal protection grounds in federal courts, so this may be an area to watch as more law enforcement agencies implement FRT.