The Detroit Police Department has agreed to new limits on facial recognition technology as part of a legal settlement, creating court-enforceable guardrails for a tool that has already led to a wrongful arrest in the city.
The settlement follows a lawsuit by Roger Williams, a Black man who was arrested after facial recognition technology identified him. Williams was represented by lawyers from the American Civil Liberties Union and the Civil Rights Litigation Initiative at the University of Michigan Law School.
What the new rules change
The central change is simple: police can no longer arrest someone based only on the output of a facial recognition search. The new policies also block arrests based only on photo lineups conducted immediately after a facial recognition search.
That matters because the policy treats facial recognition as an investigative lead, not as enough evidence by itself. Under the rules, a photo lineup cannot be conducted solely because facial recognition produced a possible match.
Instead, police must have additional evidence connecting a suspect to the crime. The settlement also requires police training on the risks and dangers of facial recognition technology.
The policies can be enforced by a court for the next four years. The police department must also audit all cases since 2017 where facial recognition was used to obtain an arrest warrant.
Why Roger Williams sued
Williams was wrongly identified as a man shown in surveillance footage stealing five watches from a store in downtown Detroit. His driver’s license photo appeared in a facial recognition search of a database containing mugshots and license photos.
A security contractor who provided the footage agreed that Williams was the best match. That identification led to his arrest.
Williams reportedly spent 30 hours in jail. Prosecutors later dropped the charges.
As part of the settlement, the police department said it is paying Williams $300,000.
“With this painful chapter of our lives closing, my wife and I will continue raising awareness about the dangers of this technology,” Williams said in a statement.
The civil rights concerns behind the settlement
The ACLU described the settlement as achieving “the nation’s strongest police department policies and practices constraining law enforcement’s use of this dangerous technology.” The group also noted that women and people of color are “substantially more likely to be misidentified by facial recognition technology.”
Those concerns sit at the center of the new Detroit policies. If facial recognition can point police toward the wrong person, then rules about what happens after a search become crucial.
The settlement draws a line between a technology-generated match and evidence strong enough to justify an arrest. It also adds review and training requirements, which are meant to address both past and future use.
The audit requirement is especially significant because it looks backward. Detroit must examine all cases since 2017 where facial recognition helped obtain an arrest warrant, not only change practices going forward.
Detroit’s policy in a broader debate
The police department, in its own statement, said it is “pleased with its work with the ACLU and University of Michigan over the last year and a half.” It also said it “firmly” believes the new policy “will serve as a national best practice and model for other agencies using this technology.”
The settlement arrives as facial recognition remains contested in law enforcement. Cities including San Francisco have banned the use of facial recognition by law enforcement. Microsoft also recently banned police departments from using its AI tech for facial recognition.
Detroit’s approach is different from a ban. The new rules still allow facial recognition technology, but they restrict how it can be used in investigations and what police may do after a match appears.
In practical terms, the settlement says a facial recognition result cannot stand alone. Police need additional evidence before moving from a search result to a lineup, and they cannot make an arrest based only on that result or on an immediate lineup built from it.
What this means for future investigations
The policy’s biggest effect may be procedural. It changes the steps police must follow before a facial recognition lead can contribute to an arrest.
The required training also acknowledges that the technology carries risks. The source of those risks is not treated as theoretical in this case: Williams’ arrest shows how a facial recognition match, a contractor’s agreement, and surveillance footage can combine into a real-world mistake.
For people concerned about police technology, the settlement offers a model based on limits, documentation, training, and court oversight. For police agencies using facial recognition, it sets out a narrower role for the tool: useful only when supported by evidence beyond the search result itself.
The case also leaves a clear public record of the stakes. Williams was arrested, spent time in jail, and later saw the charges dropped. The new Detroit rules are designed to reduce the chance that a facial recognition search alone can drive that chain of events again.