Robert Dillon spent a night in a cold cell. He rode in a caged van without lights. All because grainy photos from a fast-food restaurant camera somehow pointed to him.
The 52-year-old commercial crabber from Fort Myers, Florida, had never set foot in Jacksonville Beach. The incident occurred more than 300 miles from his home. Yet in August 2024, Jacksonville Beach police arrested him for attempting to lure a child at a McDonald’s. The sole link? A facial recognition match with 93 percent confidence.
Now Dillon is fighting back. On June 10, 2026, the American Civil Liberties Union filed suit on his behalf against the City of Jacksonville Beach, the Jacksonville Sheriff’s Office, the Pinellas County Sheriff’s Office, and individual officers. The complaint details how investigators ignored clear evidence of his innocence while building a case on shaky technology.
Officers responded to the McDonald’s after a report of a man trying to persuade a girl younger than 12 to leave with him. They reviewed security footage but did not seize it. Instead, one officer photographed the screen with a cellphone. The resulting images showed low resolution. The suspect’s face appeared partially shadowed and off-axis.
An investigator from the Jacksonville Sheriff’s Office fed those cellphone photos into the Face Analysis Comparison and Examination System, known as FACES or FACESNXT. Operated by the Pinellas County Sheriff’s Office and shared with other agencies, the system returned Robert Dillon as a possible match. Police reports cited the 93 percent figure. That became the foundation for an arrest warrant.
But problems piled up fast. Dillon’s alibi placed him far away. License-plate readers found no trace of his vehicle near the scene. He had no connection to Jacksonville Beach. None of that stopped the process. Detectives created a photo lineup including Dillon’s image. A witness picked him. The ACLU argues the lineup was tainted. When facial recognition produces a false positive, it often selects someone who merely resembles the actual suspect. That resemblance can sway witnesses.
“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 local news, as reported by Slashdot. Waters himself is among those named in the suit because his office ran the search.
The ACLU’s filing pulls no punches. It accuses authorities of concealing exculpatory evidence. They obtained the warrant without disclosing the poor image quality or the distance issues. Dillon was arrested at his home in front of his wife. The experience left scars. “I feel like people are saying, hey, there’s that guy that was on the news, stay away from him,” he told CBS News.
This case is not isolated. The ACLU counts Dillon as one of at least 15 known individuals in the United States wrongfully arrested due to facial recognition errors. The technology’s accuracy drops sharply with low-quality probe images. Shadows, angles, and resolution matter. Here the input was a photo of a screen. Experts have warned for years that such matches should never stand alone as probable cause.
Pinellas County Sheriff Bob Gualtieri oversees the FACES system. His agency leases it to other departments across Florida. The tool ranks among the oldest police facial recognition platforms in the country. Its continued use despite documented failures raises questions about oversight and training. Officers appear to have treated the 93 percent score as near-certain identification. They pressed forward even as contradictions mounted.
And the human cost shows. Dillon endured overnight detention. He faced public suspicion tied to a heinous accusation. The charges were eventually dropped once the mismatch became undeniable. Yet the damage lingers. His lawsuit seeks accountability for the agencies and individuals involved. It argues that reliance on flawed technology, paired with disregard for alibi evidence, violated his constitutional rights.
Recent coverage highlights broader concerns. WIRED examined how this incident tests one of the nation’s longest-running police face recognition tools. The report details how the system’s operators presented the match without sufficient caveats. It also notes the transportation in an unlit, caged van – a stark reminder of the physical realities of arrest.
The Guardian reported on the suit’s filing, stressing that Dillon was prosecuted despite living hundreds of miles away. The piece quotes the ACLU stating that the investigation “resulted in the wrongful arrest and prosecution of an innocent man.”
State and local law enforcement have embraced facial recognition for its speed. Proponents argue it helps solve cases and identify suspects quickly. But cases like Dillon’s reveal the gaps. Poor inputs produce poor outputs. Confirmation bias can lead investigators to overlook inconsistencies. When the accusation involves a child, the pressure to act intensifies. That pressure should not override basic investigative standards.
So what happens next? The lawsuit will test whether courts hold agencies responsible for overreliance on this technology. It may push departments to adopt stricter protocols. Require corroborating evidence. Demand higher-quality images. Train officers on the limitations of probabilistic matches. Without such changes, more innocent people risk similar ordeals.
Dillon’s story spreads on social media. Users on X call the arrest outrageous. They question how probable cause was established on a single algorithmic suggestion. The conversation reflects growing skepticism about unchecked deployment of these systems in policing.
Facial recognition has improved in some respects. Yet error rates persist, especially across demographic groups or with challenging footage. Law enforcement agencies often keep their specific performance metrics private. That secrecy complicates public assessment of the risks. Dillon’s case offers a concrete example. A commercial crabber. A cellphone snapshot. A 93 percent claim that unraveled under scrutiny.
The suit, formally titled Dillon v. City of Jacksonville Beach, was prepared with the ACLU of Florida and the firm Hoguet Newman Regal & Kenney. It seeks damages and reforms. Its success could influence practices far beyond one Florida county. Other departments using shared systems like FACES may reconsider how they document and justify matches.
Meanwhile, the girl at the McDonald’s remains safe. The real perpetrator was never identified through this process. That fact stings. Resources spent pursuing Dillon could have gone elsewhere. The technology that promised efficiency instead created a new form of error. One with human faces and real consequences.
Authorities defend their actions by pointing to the witness identification. But the ACLU counters that the witness was shown a lineup shaped by the faulty match. Such circularity undermines the entire chain. Independent verification should have been mandatory. Distance, alibis, and lack of physical evidence demanded it.
As the case proceeds, it will force a closer look at the contracts and policies governing facial recognition databases. Who bears responsibility when the system errs? The vendor? The operator? The arresting agency? Dillon’s attorneys aim to establish clear lines of accountability.
For now, Robert Dillon tries to rebuild. The news coverage has faded for most. For him the memory stays fresh. A night in custody. Suspicion from neighbors. The knowledge that an algorithm, fed bad data, upended his life. His lawsuit stands as a warning. Technology assists investigations. It must not replace judgment.
WebProNews is an iEntry Publication