Six federal agencies used facial recognition software to identify protesters who demonstrated following George Floyd's death at the hands of Minneapolis police last year, according to a government watchdog.
The Government Accountability Office (GAO) said in a report released Tuesday that the agencies used the technology from May through August 2020 to “support criminal investigations related to civil unrest, riots, or protests.”
“All six agencies reported that these searches were on images of individuals suspected of violating the law,” the GAO said.
The Bureau of Alcohol, Tobacco, Firearms and Explosives, the U.S. Capitol Police, the FBI, the U.S. Marshals Service, the U.S. Park Police and the U.S. Postal Inspection Service reported using the technology.
The use of facial recognition came under scrutiny last year over concerns about how state and local law enforcement was using the technology to identify protesters.
Facial recognition technology has long been criticized for misidentifying women and minorities at a higher rate. In April, a man sued the city of Detroit for being falsely arrested for after his driver’s license photo was mistakenly matched to surveillance footage of a shoplifter.
Last month, officials announced that the National Capital Region Facial Recognition Investigative Leads System — which was used to identify demonstrators in Lafayette Square last year — will end on July 1.
IBM and Microsoft suspended the sale of facial recognition to law enforcement last summer following the Floyd protests. Amazon in April extended its ban on selling facial recognition technology to police until further notice.
The GAO’s report found that the use of facial recognition technology was common across the federal government.
In total, 20 federal agencies reported owning a system with facial recognition technology or using another entity’s system. Only three agencies used a system that they owned, 12 agencies used another entity's system, and five agencies used systems they owned as well as another entity's system.
However, 13 agencies were not aware of what nonfederal systems were being used by employees, meaning that the agencies had not fully assessed the risks related to such systems, including privacy and accuracy.