Detroit police have been using facial recognition software for more than a year under standard operating guidelines, but without a formal policy approved by the Board of Police Commissioners. That’s not OK. The software has tremendous potential for violating civil liberties, and its use should come with restrictions that are clearly understood by both the public and the police.

The board has delayed voting on formal procedures, insisting citizens can trust police to use the technology properly to assure a safer community.

But any intrusive measure employed by law enforcement should come with strong oversight. Along with privacy concerns, the software could also run afoul of Fourth Amendment protections against illegal search and seizure.

The facial recognition initiative is separate from the city’s Project Green Light, in which business owners can purchase cameras that provide a live video connection to the police department. However, footage from the cameras may be scanned by the facial recognition software.

Many business owners and residents have applauded the Green Light program for keeping the city safer. Indeed, the cameras have helped solve crimes, says Willie Bell, a district chairman on the Police Commissioners’ board. Footage from the cameras has helped track down suspects in the May shooting of five members of the LGBT community, as well as the suspects who halted traffic on Lodge Freeway while performing stunts in their vehicles.

Yet others question the effectiveness of the program. Eric C. Williams, an attorney with ACLU of Michigan, says there is no evidence to suggest that either Project Green Light or use of facial recognition technology has helped prevent crime.

“Crime trends in Detroit have been falling parallel to the national trend,” Williams said. “Our line is just higher.”

Further, Williams expressed concern that facial recognition is especially problematic for minorities, since it has proven less accurate at recognizing black and brown faces.

“The law has not caught up to technology, and rules haven’t been made to address our fundamental rights,” Williams says.

Bell assures that any problems with the technology will be addressed, and that the department are taking constitutional concerns seriously.

A plan signed in April said the software cannot be used “unless there is reasonable suspicion that such use of facial recognition technology will provide information relevant to an active or ongoing criminal or homeland security investigation.”

But there are no regulations on constitutes “reasonable suspicion.” Judgment in this case is based on subjective evaluations or on algorithms — which some studies have shown to be inaccurate.

Though the police say the facial recognition software is used only to identify violent criminal offenders, new technology often opens a Pandora’s box. One potential problem, Williams says, is the department’s potential to share its database with other law enforcement agencies, including federal immigration officials.

“It’s not even just a matter of if you trust any individual on the Detroit Police Department,” Williams says. “It’s whether or not you trust the U.S. government.”

The Detroit Police Department must address these concerns in producing a comprehensive and clear policy for who, when and why facial recognition technology is used. And since the program is already in place, those guidelines should come as soon as possible.

— The Detroit News