(Ed. Note: The “Innovation Fact of the Week” appears as a regular feature in each edition of ITIF’s weekly email newsletter. Sign up today.)
The California State Senate is considering a bill to make it illegal for law enforcement to use facial recognition technology in officer-worn body cameras. The bill, AB 1215, has passed the State Assembly. It should not pass the Senate because it is based on a number of fallacies concerning facial recognition technology and would undermine beneficial uses of it by law enforcement.
Facial recognition technology matches an image of a face to other images in a database. To do so, the technology maps features of the face, such as the distance between the eyes, to create a vector, or “faceprint” of a face. When used with body cameras, facial recognition can help law enforcement quickly identify individuals. For example, facial recognition could help police identify people they are on alert for at a public event or determine the true identity of someone presenting false identification during a lawful stop. It could also help identify victims of human trafficking, lost or abducted children, and missing elderly individuals. The proposed law would eliminate all these benefits and more.
Rather than ban the technology, policymakers should create guardrails to prevent potential abuses. One concern voiced is that incorporating facial recognition in body cameras would transform the cameras from tools of accountability to “roving surveillance systems,” possibly chilling free speech in public places. But policymakers can easily address this concern. First, no police department in California has sought to build a permanent database of information gleaned from body cameras. California law already requires that police keep video recorded by a body-worn camera for at least 60 days, and policymakers should in addition require that law enforcement delete faceprints after a specific period of time if the data is not relevant to a case. Second, policymakers can require that law enforcement only match faces against a database of specific groups of individuals, such as lost children or individuals with outstanding arrest warrants. Third, lawmakers can establish conditions under which law enforcement can use facial recognition systems to analyze existing body camera images already in the system.
Another concern is that facial recognition technology will negatively impact women and people of color by providing false matches. The ACLU amplifies these concerns by publishing misleading results, including a recent claim that the images of many California lawmakers were mismatched against a mugshot database. However, the ACLU set artificially low confidence thresholds for matching faces to create these results so that many false positives were allowed, creating the impression that facial recognition technology is inaccurate. But these systems are accurate. The National Institute of Standards and Technology has done extensive testing of facial recognition algorithms and in 2018 it found the best algorithms had error rates below 0.2 percent.
Instead of banning uses of the technology, policymakers concerned about false matches should require law enforcement use high confidence thresholds to limit false positives to a very low level. Moreover, even when there are false matches in the field—whether caused by a computer or human error—law enforcement officials use multiple methods to verify someone’s identity and they require additional evidence to arrest and prosecute individuals.
Lastly, facial recognition technology in body cameras can increase police accountability. For example, the technology can decrease the number of unlawful stop-and-frisks. Police officers will have less of a pretext to stop someone if that person’s face does not match the face of someone with an outstanding arrest warrant. This can also help prevent racial profiling, which can result in over-policing of people of color and unlawful stops of people suspected of being undocumented.
Banning facial recognition in body cameras would take a useful tool out of the hands of law enforcement. A better approach is to adopt rules that address legitimate concerns while enabling police to use modern technology to operate efficiently, safely and in the interest of our communities.