IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Lawsuit Highlights Legal Risks of Facial Recognition Tech

Multiple Minnesota law enforcement agencies face a civil rights lawsuit over the use of facial recognition technology in an arrest. However, the government denies facial recognition led to the arrest.

facial
A Minnesota man has filed a federal civil rights lawsuit against Minnesota's Hennepin County, the city of Bloomington and multiple law enforcement employees within the two agencies. The lawsuit argues faulty facial recognition technology played a role in the violation of his rights, leading to his arrest, imprisonment and prosecution for a crime he did not commit.

Kylese Perryman was arrested and accused of carjacking and armed robbery in 2021. According to the lawsuit, he spent five days in jail and was prosecuted for the crime for 52 days before the charges against him were dropped. Perryman is a Black man, and multiple studies have shown poorer facial recognition accuracy rates for subjects who are female, Black and 18-30 years old.

However, there’s a big question about the role facial recognition technology actually played in the criminal prosecution, as Hennepin County denies it was used to produce a lead in the case.

Before the federal civil rights lawsuit was filed on June 28 of this year, Hennepin County Sheriff’s Office Analyst Nicole Hughes testified in a legislative subcommittee meeting in November 2021 about the role Hennepin County’s Criminal Information Sharing and Analysis Unit (CISA) played in the investigation.

“Facial recognition was attempted, and there were no strong leads developed from it,” testified Hughes. “It was not facial recognition that led to that person’s name. It was going through an unrelated violent crime in another city reviewing parties involved and booking photos where it does look similar.”

The Hennepin County Sheriff’s Office provided KMSP FOX 9a statement in 2021 that further detailed how Perryman’s name was provided to Bloomington Police.

"Bloomington PD asked CISA to utilize Investigative Imaging Technology (IIT) on a surveillance video to see if they were able to identify a lead on a possible suspect. The photo from the surveillance video was clear, however not high resolution. CISA was unable to provide a lead using IIT and advised Bloomington PD," the statement reads. "Later that day, the same CISA analyst was investigating an unrelated active case. The CISA analyst was reviewing booking photographs and surveillance photographs. Among those booking photographs was that of Kylese Perryman. The analyst recognized Perryman’s booking photograph and saw that it appeared to resemble the suspect from the surveillance footage in the Bloomington case. CISA sent Bloomington PD Perryman’s name, date of birth and booking photo.”

During the Subcommittee on Data Practices hearing, Hughes also explained that the agency’s policy is to only use facial recognition technology to compare suspect photos with booking photos. She stated if a match is found, she’ll proceed by checking the incarceration status of the subject at the time the suspect photo was taken, compare criminal history and potentially check in with a parole officer who has frequent contact with the party and might be able to recognize the suspect.

“Most importantly, we work to rule out a lead, not to rule them in,” Hughes said to the committee. “The responsibility still falls on the investigator to develop or rule out leads. Facial recognition technology should never be the sole means of identifying a suspect.”

According to Perryman’s lawsuit, filed by the American Civil Liberties Union of Minnesota and pro bono attorneys at Nelson Mullins Riley & Scarborough LLP, the defendants “breached their own policy by relying on facial recognition software to positively identify Kylese Perryman and ignoring all the known differences between him and the true suspect.”

Perryman's lawyers are asserting in the case that the suspect in the surveillance videos that was compared to Perryman's booking photo from a prior criminal offense had visual differences in height, weight, tattoos and hairstyle.

“The defendants have been deliberately vague and a moving target of sorts as it relates to when and whether facial recognition technology was used,” said Nelson Mullins attorney Claire Barlow in a statement to Government Technology. “In Hennepin County’s statement following the expungement, they stated that facial recognition was initially used by Hennepin County’s CISA analysts, but the images were not clear enough to proceed. They continued by stating that the same day as the attempted use of facial recognition technology, CISA analysts and/or Bloomington Police Detective Andrew Risdall personally identified Kylese Perryman by viewing previous booking photos. Hennepin County also represented to the court during Mr. Perryman’s expungement process that he was identified by facial recognition software.

"The lawsuit will help us determine which story as to the means of identification is true. Regardless of which facial recognition technique was used — software, human review or some combination of the two — the city and county law enforcement completely failed to then take even basic investigative steps like considering height, weight and tattoos, which would have ruled Mr. Perryman out. They also failed to do a lineup, follow up with witnesses or check Mr. Perryman’s alibis that show he was nowhere near the crimes. Instead, they arrested, charged and jailed an innocent man.”

The suit argues that Bloomington and Hennepin County law enforcement violated Perryman’s Fourth Amendment protection against false arrest and false imprisonment, and failed to train and supervise law enforcement on proper identification and arrests.

Perryman is now seeking $250,000 in damages.
Nikki Davidson is a data reporter for Government Technology. She’s covered government and technology news as a video, newspaper, magazine and digital journalist for media outlets across the country. She’s based in Monterey, Calif.