Police in Florida arrested a man and held him in jail for nearly three months based on a bad facial recognition result. It’s far from the first time such a thing has happened—in fact, it’s at least the second time that the same sheriff’s office was involved, according to Action News Jax.
In April 2025, a man in Jacksonville, Florida, purchased a car from someone he met in a grocery store parking lot. When he learned the car was stolen, he reported the crime to the Jacksonville Sheriff’s Office. Using surveillance footage from the parking lot, investigators ran the suspect through facial recognition software, which flagged Jalil Richardson as an 85 percent match.
The case had problems. First and foremost, Richardson lived in Charlotte, North Carolina—400 miles away. Not only that, but time cards showed Richardson was at work when the suspect was selling a stolen car in Jacksonville.
But that didn’t dissuade police. Richardson was arrested in North Carolina and held for 33 days, at which point he was extradited to Jacksonville and held for another 53 days. Prosecutors finally dropped the charges and released him last month after he had spent nearly three months in jail.
Richardson told Action News Jax that the arrest and confinement cost him his job, his house, and custody of two of his children.
Arresting someone for a crime he did not commit is among the worst things law enforcement can do, especially if that arrest causes negative results in the arrestee’s life. But this is not the only time in recent memory that cops in Jacksonville have made this mistake.
Earlier this month, the American Civil Liberties Union (ACLU) sued the city of Jacksonville Beach, a separate municipality just east of Jacksonville. Police officers were investigating an attempted child abduction when a Jacksonville Sheriff’s Office investigator ran photos of grainy surveillance footage of the suspect through facial recognition. The software flagged Robert Dillon, whom officers arrested and charged, even though he lived and worked 300 miles away.
The ACLU identified at least 15 people, including Dillon and Richardson, who had been arrested since 2019 based on bad facial recognition.
“Facial recognition software is just one tool in a large toolbox for investigators,” the Jacksonville Sheriff’s Office said in a statement to Action News Jax about Richardson’s case. “Our detectives and officers use any and all available resources to solve cases. It is incorrect to assume that facial recognition was the deciding factor in Mr. Richardson’s arrest.”
“If you came to me with a facial recognition hit and that was your probable cause, I would probably kick you out of my office because that’s not how it works,” Jacksonville Sheriff T.K. Waters told Action News Jax last year regarding Dillon’s case. “There better be a lot more that goes along with that to help make sure that we have the proper individual too.”
But in Richardson’s case, the only other investigative technique identified was a photo lineup, which investigators administered to both the witness and his brother.
As the ACLU noted in its lawsuit on Dillon’s behalf, when facial recognition “produces a false match, the returned (innocent) candidate will, by algorithmic design, resemble the actual perpetrator. Placing that candidate’s photograph in an array predictably taints any witness identification that follows. Photo arrays are constructed by surrounding the candidate with ‘fillers’—photographs of known innocents selected for their physical similarity to the candidate, not to the actual perpetrator.”
This is not to say that facial recognition has no use as an investigative tool, but it’s clear that it should not form the sole, or perhaps even primary, basis for identifying a suspect.
Then again, perhaps even that is insufficient: “The technology is simply too dangerous for law enforcement to be using at all,” Adam Schwartz, privacy litigation director at the Electronic Frontier Foundation, told Action News Jax.
