A Florida man says his life changed because a computer recognized a face, and police treated that result as an answer.
Robert Dillon was not at the Jacksonville Beach McDonald’s where police were investigating an alleged attempt to lure a child. He lived more than 300 miles away, had visible facial scars, he says did not match the suspect, and later saw the charges against him dropped. Still, according to a new lawsuit backed by the ACLU, he was arrested after facial recognition software produced a 93 percent similarity result that placed him in the center of a case he says had nothing to do with him.
We should not read this as a strange one-time technology failure. We should read it as a warning about what can happen when an AI-generated lead walks into a police investigation and starts wearing the clothes of proof. Dillon’s lawsuit against Jacksonville Beach officials and others puts one sharp question in front of the public: when facial recognition gets a person wrong, who pays the price?
A 93 Percent AI Match Became the Start of a Nightmare
The case began with a frightening report at a McDonald’s in Jacksonville Beach, Florida. Police were told that a man had allegedly approached a girl and tried to get her to leave with him. Officers had a surveillance image, but they did not have a confirmed name.
That is where facial recognition entered the story. According to the lawsuit and public reporting, investigators used FACESNXT, a Face Analysis Comparison and Examination System, to compare an image of the suspect with law enforcement records. The system returned Dillon as a 93 percent match on facial features.
That number sounds powerful to the public. It sounds scientific, clean, and almost final. But a facial recognition score is not the same thing as a fingerprint, a confession, a location record, or an eyewitness account from the person directly involved. It is a machine-generated lead based on visual similarity, and its strength depends heavily on image quality, camera angle, lighting, database size, algorithm design, and human interpretation.
Dillon’s lawsuit argues that police allowed that AI result to take over the investigation. Instead of treating the technology as a clue that needed to be tested, the complaint says officers built the case around it.
What the Lawsuit Says Police Missed
Dillon says he told police he was nowhere near Jacksonville Beach. He lived hundreds of miles away, and according to the complaint, there were obvious reasons to question whether he could have been the man in the surveillance image.
The lawsuit says the suspect image was not a clean digital frame pulled directly from original surveillance footage. Instead, it was allegedly captured from a computer screen using a cellphone. That detail matters because facial recognition systems are sensitive to the quality of the input image. A grainy, shadowed, off-angle image can make two different faces look more alike than they would in a clearer comparison.
Dillon also says he has distinctive scars from skin cancer surgery running from his hairline toward his nose. He says the scars visible in his own photo did not match the suspect’s appearance. In a traditional investigation, that kind of visible difference should raise a loud red flag.
The lawsuit also claims police had information that could have pointed away from Dillon. It alleges that license plate reader checks did not place his vehicles near the restaurant and that a McDonald’s employee reportedly described the suspect as a regular customer. If Dillon had never been to Jacksonville Beach, the fact that he was a familiar customer there should have warranted deeper scrutiny.
The ACLU’s central claim is direct: police did not simply use facial recognition AI. They allegedly relied on it to the point of replacing basic investigative work.
Why the Lawsuit Could Matter Beyond Jacksonville Beach
Dillon’s lawsuit is not only about damages for one man. It is also about whether courts will force clearer rules around police use of AI-driven identification tools.
The lawsuit seeks accountability and policy changes. If it succeeds, it could pressure agencies to create stronger safeguards around image quality, documentation, lineup procedures, warrant disclosures, and independent corroboration. Even if the case does not create a national standard, it can still add to the growing legal pressure around facial recognition policing.
That pressure is already building. Cities and states have taken different approaches, from restrictions to bans to policy reforms. Some law enforcement agencies argue that facial recognition remains valuable when used properly. Civil liberties groups argue that the harms are too severe without strict rules and transparency.
Both points can be true at once. A tool can help solve crimes and still be too dangerous when used carelessly. The public safety benefit does not erase the constitutional risk. The constitutional risk does not erase the need to identify dangerous suspects. The real debate is whether police departments can prove they have built a system that prevents a lead from becoming a life-altering mistake.
Read the original story on Crafting Your Home
