Mistaken arrest in Georgia triggered by false facial recognition match in different state
Racial bias in facial recognition is getting a close-up, after police in Georgia used facial biometrics to arrest a man on a fugitive warrant from a state he’d never been to, AP reports.
Twenty-eight-year-old Randal Reid was detained in DeKalb County, Georgia, in late November 2022, after authorities in Louisiana used facial recognition algorithms to erroneously link him to a purse theft incident in Baton Rouge.
“They told me I had a warrant out of Jefferson Parish,” Reid told the Times-Picayune. “I said, ‘what is Jefferson Parish?’ I have never been to Louisiana.”
Reid was released after five days. But the case raises fresh concerns about the accuracy of facial recognition in use by law enforcement, who has jurisdictional authority over biometric data, and the use of safeguards. Reid reportedly had a mole on his face that was not recognized by FRT. His attorney, Tommy Calogero, estimated a difference of 40 pounds between Reid and the figure captured by surveillance cameras, to whom he was linked.
Official laws say face recognition tools can only be used to generate leads, and only after the filing of a vetted formal request, measures meant to curb misuse. But Reid was arrested after a botched ID in Jefferson Parish prompted an official in Baton Rogue to issue a warrant, which then triangulated with authorities in Georgia — a case of broken telephone, but with facial recognition data.
The use of face scans and AI-assisted facial recognition software by law enforcement has been hotly debated in the U.S., and intersected with conversations about racial injustice, with studies showing that FRT is more likely to misidentify Black people and people of colour. And there have been previous cases of mistaken arrests resulting partly from the use of facial recognition.
These concerns over misconduct come in the wake of a pandemic in which authorities used information collected as part of the global response to COVID-19, to ramp up surveillance and other security measures.
Despite this, a 2022 Pew Research Centre report found that Americans were more likely to see facial recognition by police as being good, rather than bad for society.
In the same report, however, 66 percent of U.S. adults said they thought facial recognition technology would be used to monitor Black and Hispanic neighbourhoods much more frequently than other neighbourhoods.