Tighter policies lead to fewer facial recognition searches for Detroit police

Police use of facial recognition technology continues to be an issue of concern globally – and pressure to put proper governance and regulations in place may be turning the biometrics use case into a niche, rather than a broadly used crimefighting tool.

In Detroit, where facial recognition has led to three cases of misidentification and subsequent lawsuits against the city, police used the technology just nine times in 2025 – a 91 percent drop from 2023, according to a report from BridgeDetroit. Three searches related to murders, three to aggravated assaults, and two to robberies.

The lawsuits are part of what’s led to fewer searches, specifically a 2024 settlement agreement, which prompted fresh changes to policy and procedure.

Facial recognition only works properly when police do

Detroit police have been using facial recognition since 2017, starting with South Carolina-based provider DataWorks Plus. The department’s first facial recognition policy was created in 2019 to protect against potential abuse. Those provisions say FRT can only be used in cases of suspected violent crimes or home invasions. Use for surveillance, specifically at “‘First Amendment events’ such as protests and other constitutionally-protected activities,” is prohibited.

More significantly, a facial recognition match can not be used as the sole basis for arrest. It can only ever provide a lead. According to a report from the Detroit Board of Police Commissioners, only one of the nine facial recognition searches conducted last year produced an investigative lead.

Race is also a major issue: all of the requests were for Black males. Concerns around FRT for law enforcement often focus on how the tech tends to perform worse when matching people with dark skin. Although the most accurate facial matching algorithms show very low differentials in recent biometric testing by the National Institute of Standards and Technology (NIST), the organization has found that a majority of facial recognition algorithms are more likely to misidentify people with darker skin, women and the elderly.

All three of the victims that facial recognition misidentified to Detroit police, leading to arrests, are black.

Robert Williams and Michael Oliver were misidentified in 2019. Detroit police used facial recognition the most in 2021, when they conducted 136 searches. By 2023, the number had fallen to 100 – one of which was Porcha Woodruff, who was eight-months pregnant when police arrested and detained her in a cell for 11 hours on suspicion of robbery and carjacking. She subsequently sued the Detroit Police Department for wrongful arrest and imprisonment.

That’s surely a factor in the significant dip that occurred in 2024, when the force used facial recognition tech just 28 times.

The story is consistent with how police use of the tech has unfolded thus far: the tech works, and with proper guidelines in place, it can be a useful tool. However, that is reliant on the assumption that police will follow the guidelines – which, in Detroit and elsewhere, has not always been the case.

FRT as ‘last resort’ raises questions about sustainable market

Perhaps the most telling statement comes from Board of Police Commissioners Member for District 7 Victoria Camille. “If it’s not being used (or) hardly at all, that’s a good thing,” Camille tells BridgeDetroit. “It’s something we really want to reserve for the last resort.”

That line of thinking does not lead to large, lucrative contracts for facial recognition providers at scale. It suggests increasing caution in using facial recognition as police come under greater scrutiny – leaving FRT as a smaller slice of the security pie than some anticipated.

Per the report, between 2017 and 2022, the Detroit Police Department spent $2,373,084.30 on facial recognition.

Since then, it has not spent a dime.

Part of the reason is that it can borrow the tools from Michigan State Police, which allows other agencies to use its facial recognition technology. Detroit Police is among those that have a Memorandum of Understanding with MSP for this purpose.

And so the questions arise: how many facial recognition systems does it take to support law enforcement in the U.S.? Should police across the country just lean on Michigan? Is FRT actually a toolkit in the police arsenal over the long term? And how do advancements in AI affect the outlook?

The answers will depend, in part, on the number of days, months or years until another misidentification triggers further legal action.

Meanwhile, facial recognition use cases continue to take root beyond law enforcement. A new facial recognition software system called Visitor Aware is now in place at every school in the Detroit Public Schools Community District. According to a Detroit News report, the system matches visitors’ faces against a valid form of identification before they can enter the school.

Visitor Aware’s system, which is used by 360 districts across the U.S., has prompted privacy concerns – but also shown that overlap in understanding of different facial recognition systems can be problematic. Although the majority of students in the district are Black, alarm over bias does not apply as it does in the law enforcement context; face-to-ID matching, or 1:1 FRT, is not the same as FRT that matches one against many (1:n).

Detroit school district spokesperson Chrystal Wilson says “we use it against an actual ID, not against a national database, so we don’t have that issue (of someone being falsely identified) because when that person’s picture is taken, it’s recognizing them against an ID and nothing more.”

Article Topics

biometric matching  |  biometrics  |  criminal ID  |  Detroit  |  facial recognition  |  law enforcement  |  schools