A 93 percent match, a 300-mile alibi, and a night in a Florida jail

A facial recognition system told Florida investigators it had found their child-luring suspect with 93 percent confidence, and the man it named was a 52-year-old retiree who had never set foot in the town where the crime happened.

The case Florida built on a similarity score

On June 10, 2026, Robert Dillon, a 52-year-old resident of Fort Myers, Florida, filed a federal civil rights lawsuit against the Jacksonville Beach Police Department, the Jacksonville Sheriff's Office, the Pinellas County Sheriff's Office, and individual officers. The case, captioned Dillon v. City of Jacksonville Beach, is being argued by the ACLU, the ACLU of Florida, and the New York firm Hoguet Newman Regal & Kenney, LLP. The complaint describes how a faulty algorithmic match became the entire investigation rather than a starting point for one.

The underlying crime was the attempted luring of a child at a McDonald's restaurant in Jacksonville Beach on November 2, 2023. Eight to nine months later, in August 2024, Jacksonville Beach Police arrested Dillon and charged him. He was held overnight, then forced to borrow money and pledge the title to his pickup truck to post bond. Prosecutors later dropped the charges after finding no supporting evidence, and his formal arrest record was wiped. His mugshot, however, still circulates online.

The geography alone should have been disqualifying. Dillon lives more than 300 miles, roughly a five-hour drive, from Jacksonville Beach, and he says he had never set foot in the town. An automatic license plate reader search later showed his truck had not been anywhere near Jacksonville Beach in the days surrounding the alleged crime. None of that information stopped the arrest warrant from being signed.

How FACES turned a grainy still into a suspect

The match came out of FACES, the Face Analysis Comparison and Examination System, a facial recognition database operated by the Pinellas County Sheriff's Office. The system has been running since 2001, making it one of the oldest law-enforcement face databases in the country. At its 2021 peak, more than 260 federal, state, and local law enforcement agencies had access to it.

According to the complaint, a Jacksonville Sheriff's Office employee ran grainy surveillance photos of the real suspect through Pinellas FACES and received a candidate match against Dillon, flagged at 93 percent on facial features. That score, by itself, was treated as enough to act on.

The lead investigator was Officer Scott O'Connell of Jacksonville Beach Police, who is now a corporal. Court filings allege O'Connell built an arrest warrant on the 93 percent match alone. The filings further allege he ignored a McDonald's employee who described the real suspect as a "regular" customer of the store, and that he left exculpatory evidence out of the warrant application. Dillon also has distinctive scars running from his hairline to his nose from skin cancer surgery, a feature the suspect in the McDonald's surveillance video did not appear to share.

O'Connell's history is part of the complaint. He had previously been terminated from the St. Johns County Sheriff's Office after threatening to "blow up" the agency, and was later arrested on a domestic battery charge before being hired at Jacksonville Beach.

A pattern, not an incident

Dillon's filing arrives in a year when the same software-as-evidence problem has surfaced repeatedly in the United States. According to the ACLU, Dillon is one of at least 15 publicly known people in the country who have been wrongfully arrested after being misidentified by police facial recognition software.

Several of those cases trace back to the same Florida database. Other Florida wrongful arrests linked to the Pinellas FACES system include Beau Burgess in New Smyrna Beach in 2025 and a separate Orlando-area case earlier in 2026. In 2024, the Jacksonville Sheriff's Office wrongfully arrested Jalil Richardson, a North Carolina resident misidentified as a vehicle thief. Richardson spent more than 50 days in a Florida jail before North Carolina work timesheets cleared him.

The nationally known earlier cases form the rest of the pattern. Robert Williams was wrongfully arrested in Detroit in 2020, Randal Reid in Louisiana in 2022, and Porcha Woodruff in Detroit in 2023, each after a facial recognition hit was treated as a finding rather than a lead. In every case, an arrest preceded the kind of corroboration that ordinarily defines a criminal investigation.

"AI says I did this, how am I going to prove that I didn't?"

Dillon's own description of the arrest, given in a CBS News interview, captures the inversion at the heart of the complaint: "AI says I did this, how am I going to prove that I didn't?"

In a statement released through the ACLU on June 10, 2026, he was blunter about the personal cost. "The night I spent in jail after they arrested me for a crime I did not commit still haunts me to this day," he said. "Over a year later, I'm still picking up the pieces of my life, all because the police relied on this dangerous technology instead of doing their jobs and actually investigating."

His lawyers framed the harm as structural. "No one should lose their freedom or be scared to leave their house because an algorithm got it wrong," said Nate Freed Wessler, Deputy Director of the ACLU Speech, Privacy, and Technology Project. Nicholas Warren, a staff attorney with the ACLU of Florida, added two lines that double as the suit's thesis: "One wrongful arrest is one too many, this should have never happened to Mr. Dillon," and "Florida's growing reliance on facial recognition technology threatens us all."

No Florida law enforcement agency has formally apologized to Dillon.

What the lawsuit is actually asking for

The complaint is not only a damages action. It demands that the Pinellas County Sheriff's Office and the Jacksonville Beach Police Department immediately adopt written policies restricting facial recognition use. That ask is the operative move. The ACLU's filing argues that the failure was not the algorithm alone but the absence of any policy requiring detectives to corroborate a software match before seeking an arrest warrant.

Read against the rest of the Florida record, the policy ask is concrete. A McDonald's employee's lead about a regular customer was available and ignored. A license plate reader search that placed Dillon's truck far from Jacksonville Beach was available and, the complaint alleges, never weighed. A visible facial scar that did not appear in the surveillance footage was available to anyone who compared two photographs side by side. None of that information was treated as decisive against a 93 percent score from a system marketed as an investigative aid.

Dillon's record has been wiped, but his mugshot persists, and the truck he pledged to make bail is still parked in a Fort Myers driveway. The lawsuit asks the courts to decide whether a similarity score, on its own, can ever be enough to take a person's freedom for a night.

Sources

• Ars Technica: Man sues Florida cops over arrest spurred by 93 percent match in facial recognition
• ACLU statement on Dillon v. City of Jacksonville Beach
• CBS News interview with Robert Dillon