United StatesPublic sector2017 – ongoingDetroit Police Department
Facial-recognition leads in criminal investigations
Detroit restricted facial-recognition investigations after Robert Williams's wrongful arrest
Problem
Detroit police used facial recognition to generate investigative leads from images of unidentified suspects.
Approach
The department's facial-recognition request log began in 2017. A false match led to Robert Williams's wrongful arrest in January 2020, according to his lawyers. A June 2024 court-enforceable settlement required independent reliable evidence before a facial-recognition lead could support a photo lineup or arrest, additional training and an audit of earlier cases. Police retained the technology, and a deputy chief said the revised practices would prevent future misidentifications.
Results
Takeaway
A computer-generated lead passed into an identification and arrest process without sufficient independent corroboration. The settlement placed new requirements on the human investigation surrounding the match.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Production and qualityFinance, risk and legal
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Vendor solution
- Evidence
- Regulator, audit or court
Sources
Source summaries
The order and attached agreement established enforceable restrictions on lineups, arrests, training and disclosure. They required an audit using records beginning in February 2017 and retained court enforcement for four years.
Williams's legal representatives described the false match and wrongful arrest in January 2020. They explained the independent-evidence requirements and retrospective audit secured in the settlement.
Police defended the tool. A deputy chief expressed confidence in the new safeguards.
Full details are in the original sources.