Detroit Extends Contract of Facial Recognition Program That Doesn’t Work

The city’s own police chief admits the technology misidentifies 96 percent of time and yet the city voted to extend its use.
September 30, 2020, 2:12pm
GettyImages-75287401

Detroit City Council voted 6-3 on Tuesday to extend the city's police department facial recognition contract, despite its near 100 percent misidentification rate and almost exclusive use on Black citizens.

The measure approves a two-year contract worth nearly $200,000 with DataWorks Plus, a South Carolina-based company, that will maintain, support, and upgrade the city's facial recognition technology. According to Detroit Police Captain Aric Tosqui, the technology has been used 106 times, made 64 matches, and was used in 12 arrests. As Motherboard has previously reported, it has been used almost exclusively on Black people.

Advertisement

Council member Raquel Castaneda-Lopez, who opposed the contract and voted no, warned the technology "has proven in multiple studies to have a racial bias and that is one of the fundamental flaws." 

At least two Black men have been wrongfully arrested by Detroit police because of the department's facial recognition technology. In a public meeting late June, Detroit Police Chief James Craig admitted that "96 percent of the time it would misidentify,” a fact that was mentioned by Councilor James Tate, who still voted in favor of the contract.

In the months since the first wrongful accusation and arrest thanks to Detroit’s facial recognition technology in TK, facial recognition has been used in cities like Miami, New York, and Pittsburgh to target and arrest individuals for protesting. A horde of federal agencies have joined the effort, including the DEA, DHS, and other unidentified authorities

In the wake of protests sparked by police brutality, law enforcement agencies have expanded their surveillance infrastructure and turned it on protesters despite a fundamental inability to accurately identify faces, but especially those of Black and brown people.

Calre Garvie, a senior associate with Georgetown's Center on Privacy & Technology, shared on Twitter that at the public meeting held over Zoom, the Detroit Police Department hinted it "might continue using the tech regardless of the vote" with the only difference being “whether the tech could be upgraded (yes vote) v. become out of date (no vote).” 

Advertisement

During the public meeting, protesters organized a caravan outside the home of Councilor Andre Spivey, who voted in favor of the contract as part of a series of protest caravans stretching back to this summer protesting the city’s use of the technology. Critics also took aim at Project Green Light, a Detroit program that allows businesses to pay $4,000 to $6,000 to allow the police to monitor their security footage in real time. Businesses participating in the program feature flashing green lights outside of their property, meant to help deter crime—despite there being no independent analyses of whether the program works. Those stores are also given Priority 1 status on police dispatches, meaning they have effectively paid extra to ensure there will be no delays when they call the police.

Combine these measures with a racially biased facial recognition technology that misidentifies nearly 100 percent of faces analyzed in the largest majority-Black city in the U.S., and it is clear why there was significant pushback to what many residents dismiss as a racist program with little evidence of positive impact.