Robert Williams wrongfully arrested after Detroit facial recognition false match
A facial recognition system falsely matched Robert Williams to surveillance footage of a store theft, leading Detroit police to wrongfully arrest and detain him for about 30 hours.
What happened
In January 2020, Detroit police arrested Robert Williams outside his home in front of his family and held him in detention for roughly 30 hours. The arrest stemmed from an investigation into a 2018 watch theft at a Shinola store in Detroit. Detroit police took a blurry, low-quality still image from the store's surveillance video and ran it through a Michigan State Police facial recognition system, which falsely matched the footage to Williams' expired driver's-license photo. Williams was not the man in the footage and was not near the store. This is the first publicly reported instance in the US of a false facial recognition match leading to a wrongful arrest. The ACLU sued the City of Detroit on Williams' behalf. The case was settled on June 28, 2024, with the city paying $300,000 and adopting policy reforms, including a requirement that facial recognition results be corroborated with independent evidence before an arrest, training on the technology's higher misidentification rates for people of color, and an audit of all cases since 2017 in which facial recognition was used to obtain arrest warrants.
What the agent did
An automated facial recognition system produced a false match between low-quality surveillance footage and Williams' driver's-license photo. That match was used to identify him as a suspect. The arrest and roughly 30-hour detention were carried out by Detroit police officers acting on the match; the automated system did not itself take the enforcement action.
The irreversible effect
Williams was arrested in front of his family and detained for approximately 30 hours based on a false identification.
Root cause
A blurry, low-quality surveillance still was run through facial recognition and returned a false match to Williams' expired driver's-license photo. Officers acted on the unverified match without corroborating it against independent evidence, and the technology carries higher misidentification rates for people of color.
How a maker-checker control would have refused it
The consequential action here, the arrest and detention, was taken by human police officers, not by an automated system, so no automated control blocked it and this framing is hypothetical. A maker-checker style control would require the facial recognition match (the maker) to be independently corroborated by other evidence before a warrant or arrest (the checker) could proceed. That is essentially the reform Detroit later adopted: face-recognition results must be confirmed with independent evidence before an arrest. Such a gate, had it existed in January 2020, could have caught the false match before Williams was arrested.
Runnable reproduction
A runnable reproduction for this entry is in progress.
Accuracy and corrections
This entry describes a publicly reported incident and is compiled from the primary sources listed above. Where an account is a legal allegation rather than an established finding, the entry labels it as such. Summaries can still contain errors. If you can document a correction, email hello@makerchecker.ai and we will review and correct it, with the change noted, within 14 days.
See it for yourself
Reading is one thing. Watch it block an agent.
One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.
Designed against the rules your auditors already enforce.