How an AI facial recognition tool led to a Florida man's wrongful arrest

A federal lawsuit is bringing to light how an artificial intelligence error led to an innocent Florida man being put in jail.

Robert Dillon was at his Fort Myers home in 2024 when police showed up to arrest him after a database flagged him as a 93% match for a crime he did not commit. He was accused of attempting to lure a child at a fast food restaurant in Jacksonville Beach — 300 miles away from him.

Charges have since been dropped, and the record has been wiped, but the trauma remains.

Now, the 52-year-old, with the help of the American Civil Liberties Union, is suing the Jacksonville Beach Police Department, as well as the Jacksonville and Pinellas County Sheriff's Offices, for damages and is demanding policy changes.

The Jacksonville Sheriff's Office said it is unable to comment due to pending litigation.

Here's what else to know.

More about Dillon's case

On "Florida Matters Live & Local," ACLU attorney Nathan Wessler said the case first started in November 2023 when an unknown man walked into a McDonald's in Jacksonville Beach and had reportedly tried to lure a young girl out of the restaurant with him. She called her parents, who contacted the police, but the man was gone by the time they got there.

"The next day, the police came and got a piece of the surveillance video, the security camera footage from McDonald's that depicted an unknown man, they sent it to some neighboring agencies, including the sheriff's office in Jacksonville, which ran it through a facial recognition technology search and that process spit out our client's photo and name as a possible match — a lead," Wessler said.

The program was the Face Analysis Comparison and Examination System. It's a centralized database maintained by the Pinellas County Sheriff's Office. The system is powered by artificial intelligence. Law enforcement uses it to attempt to identify an unknown person by searching a large image repository, according to the complaint. The accuracy depends on the quality of the probe image.

ALSO READ: With no federal facial recognition law, states rush to fill void

"Lower quality images contain less interpretable facial data, degrading the system's ability to produce a reliable template," the complaint reads.

The lawsuit writes that the system returns a list of possible matches. But if the actual suspect is not in the database, "all returned candidates are necessarily incorrect, making the false positive rate 100%."

In terms of Dillon's case, law enforcement obtained an arrest warrant by having him as a possible match and a statement from a restaurant employee who picked his photo out of a lineup.

Wessler argues law enforcement "ignored evidence" that led to the arrest. According to the ACLU, a McDonald's employee said the suspect was a "regular" at the restaurant. But Dillon lived five hours away and told police he had never been to Jacksonville Beach in his life. An automatic license plate reader search also showed no hits on his car anywhere near the area around the crime's timeline.

In addition, police had called Dillon in 2023 about the crime, where he denied any involvement.

"You've got the wrong guy. I have very distinctive scars. I've had skin cancer and have a scar running from my hairline down to my nose," he told the officer.

Dillon also told the officer he would cooperate with law enforcement, and when he called Fort Myers Police, an employee said the call sounded like a hoax. When he contacted police in Jacksonville again, the person who answered told him they weren't aware of the investigation and that it sounded like a scam, according to the complaint.

He did not have any more communication about this until a Lee County sheriff's deputy showed up at his home to arrest him months later in August 2024.

According to the ACLU, Dillon spent the night in jail. He then had to borrow money and pledge the title to his truck so he could post bond. He also had to find and pay for a defense attorney. Wessler said Dillon had been working as a self-employed commercial crabber and had lost a lot of income when he couldn't work.

"He was afraid to go out in the community; he was afraid to be seen, afraid someone might start a fight with him. So he actually didn't work for more than a month, got behind on his rent, and he actually still to this day has not caught up," Wessler said.

In a prepared statement, Dillon detailed how he's still recovering.

"The night I spent in jail after they arrested me for a crime I did not commit still haunts me to this day. I will never get over how terrified and worried if I'd ever go home to my wife and daughter again," he said. "Over a year later, I'm still picking up the pieces of my life because the police relied on this dangerous technology instead of doing their jobs and actually investigating. Florida police must implement safeguards and ensure this never happens to anyone else, because until they do, nobody is safe."

The state attorney's office dropped all the charges in October 2024.

According to the complaint, Jacksonville Beach Police said all warrant requests are submitted to the state attorney's office, and it makes the decision whether to move forward. The department said it would not comment on the matter beyond that.

Worries over AI in policing

But Dillon is not the other person who has experienced this. He is one of 15 known people in the country for this to happen to them.

According to the ACLU, besides Florida, police in Maryland, Michigan, Missouri, Louisiana, Nevada, New Jersey, New York, North Dakota and Arizona have been publicly known to have wrongfully arrested people due to relying on this technology.

Wessler said in Dillon's case, it seems to "be blame up the chain here." He argues there was a lack of sufficient training for officers and for departments. He also said there's a lack of technical safeguards against low-quality input photos getting sent. He said the image used was a still from a "pretty low resolution security camera footage that was at an angle," and the face was shadowed.

ALSO READ: With no federal facial recognition law, states rush to fill void

But the Pinellas County Sheriff's Office disputes this. It says that proper safeguards were put in place to require the system to be properly used. It said it was wrong to determine probable cause existed based solely on the software, and the liability rests solely on the officer who made the decision.

"The assertion in the lawsuit that PCSO failed to train [officers in the facial recognition technology] is patently false. Facial recognition results are never matches," the sheriff's office wrote in part. "Independent investigation is required to determine whether any person in a photo array return in the facial recognition system is the person who committed a crime requires a law enforcement officer to determine probable cause through independent means."

According to the ACLU, testing of facial recognition systems has exhibited higher rates of false matches when used on people of color, women, older people and young people.

"There are members of certain communities in Florida who are just more at risk of being misidentified, but as Mr. Dillon's ordeal shows, nobody is safe," Wessler said. "Mr. Dillon is white and this technology misidentified him, flagged a totally innocent man who was not a match to the suspect and police ran with it, and completely turned his life inside out."

The complaint alleges that even after charges were dropped against Dillon, the Jacksonville Sheriff's Office "took no corrective action" against the officer who handled the search, conducted no review of its facial recognition practices and did not adopt policies to prevent recurrence.

"As Mr. Dillon often says, he wouldn't wish this ordeal on his worst enemy. There are very concrete things that police can do — things that other police departments have done after we've sued them or after they've seen some settlements right and realized there are best practices," Wessler said. "Things police departments can do to shape up to avoid the worst effects of relying on this bad technology, so we are seeking some pretty straightforward but powerful policy changes in the cities that we've sued."

This story was compiled from an interview by Matthew Peddie for "Florida Matters Live & Local." You can listen to the full episode here.