JACKSONVILLE, Fla. — The Jacksonville Sheriff’s Office is facing renewed questions over its use of artificial intelligence facial recognition technology after a second man said the system wrongly identified him as a suspect, leading to three months behind bars.
Only on Action News Jax, Jalil Richardson, a Charlotte, North Carolina father of 10, said he was extradited to Jacksonville and spent nearly three months in jail for a car theft investigators later determined he did not commit.
>>> STREAM ACTION NEWS JAX LIVE <<<
“It’s overwhelming and it’s devastating and it’s outrageous,” Richardson told Action News Jax.
According to a Jacksonville police report, investigators used Automated Facial Recognition, or AFR, to compare surveillance footage from a Publix parking lot theft case to Richardson’s photo.
Jasmine Jackson, Richardson’s wife, said officers told them the software identified her husband as an “85% match.”
“He said, ‘Jalil came back as close as 85% and that’s the reason why he charged Jalil with the crime,’” Jackson said.
The investigation began April 2, 2025, when a victim told JSO he unknowingly purchased a stolen car from a man he met at a Publix on Baymeadows Road. Police later showed the victim a photo lineup, where he identified Richardson as the suspect.
But timecards later proved Richardson was at work in North Carolina when the alleged crime happened nearly 400 miles away in Jacksonville.
“None of the people that she had even came close to looking like my husband,” Jackson said.
Richardson said he first learned about the warrant after calling police to his home in Charlotte for an unrelated disturbance.
“When they arrived, they informed me they had a warrant for me out in Jacksonville and I was incarcerated for 33 days in Mecklenburg County,” Richardson said.
“They extradited me to Florida after 33 days in Mecklenburg County and then I spent the rest of the month in Jacksonville, which was another 50 days,” he said.
Court records show prosecutors dropped the case Wednesday.
Charges were dismissed, including dealing in stolen property, grand theft, unlawful possession of motor vehicle with vehicle identification number removed, possession of fraudulent title, and fraudulent use of fictitious personal identification information.
Criminal defense attorney Michelle Suskauer, who is not involved in the case, said the technology can create serious risks when paired with poor-quality surveillance images.
“There are so many levels of error here,” Suskauer said. “It’s pretty terrifying when someone’s life can be ruined as a result of the technology used in this case.”
Action News Jax first reported concerns surrounding JSO’s facial recognition technology in July 2025 after a Lee County man was wrongfully arrested in a separate case.
[DOWNLOAD: Free Action News Jax app for alerts as news breaks]
In that investigation, Jacksonville Beach police worked with JSO and identified Robert Dillon as a 93% match to a suspect accused of attempting to lure a 12-year-old child in Jacksonville Beach in 2023. That case was ultimately dropped.
Now, Richardson’s family said the fallout from his arrest has devastated their lives.
“The case is just dismissed, but we’re homeless,” Jackson said. “He’s lost his job. He has health issues. Our kids are scattered in different houses. We lost our car. We lost everything because my husband lost his income.”
[SIGN UP: Action News Jax Daily Headlines Newsletter]
Action News Jax asked both JSO and the State Attorney’s Office whether there have been other cases dismissed because of inaccurate facial recognition matches and what percentage threshold investigators use to determine a reliable match.
JSO said it is working to answer those questions. Action News Jax is still waiting to hear back from the State Attorney’s Office.
For now, the family said they are staying with friends while trying to rebuild their lives.
Click here to download the free Action News Jax news and weather apps, click here to download the Action News Jax Now app for your smart TV and click here to stream Action News Jax live.
