A man who spent 22 days in jail because of a flawed facial recognition match has filed a federal lawsuit claiming that Florida law enforcement officers disregarded multiple forms of evidence that pointed away from him as the person responsible for a string of thefts. The case, reported by Ars Technica, highlights persistent problems with the accuracy of facial recognition technology when used by police and the potential consequences when officers treat algorithmic matches as conclusive proof rather than one investigative lead among many.
The plaintiff, Richard Williams, was arrested in 2024 after detectives in Palm Beach County linked him to a series of package thefts from residential porches. According to the complaint, the investigation began when a homeowner reported seeing someone take a delivered parcel from her front steps. The resident provided surveillance video to the sheriff’s office that showed a man wearing a baseball cap, sunglasses, and a mask that covered much of his face. Rather than relying solely on that limited visual evidence, detectives ran a still image extracted from the video through a facial recognition system operated by the Palm Beach County Sheriff’s Office.
The system returned a possible match to Williams, who had a prior criminal record. Investigators then obtained a driver’s license photo of Williams and showed it to the homeowner in a photo lineup. The resident identified him as the person she had seen. On that basis, officers obtained an arrest warrant. Williams was taken into custody in early 2024 and spent more than three weeks behind bars before prosecutors dropped the charges. The dismissal came after Williams’ defense attorney presented security footage from his workplace showing that he was on the clock and inside a building several miles away at the exact time the theft occurred.
Court records and the lawsuit allege that police possessed this exculpatory evidence early in the process but failed to follow up on it adequately. Williams’ time card, GPS data from his employer’s vehicle tracking system, and statements from coworkers all placed him far from the scene. Despite these contradictions, detectives moved forward with the arrest, treating the facial recognition hit as sufficient corroboration. The lawsuit contends that such an approach reflects a systemic overreliance on biometric software that has documented error rates, particularly when the source image is low quality or the subject is partially obscured.
Facial recognition technology has become a common tool for law enforcement agencies across the United States. Many departments subscribe to services offered by private vendors or use systems developed by federal agencies. These programs compare probe images against massive databases that can include driver’s license photos, mug shots, and even images scraped from social media. While the software can return ranked lists of potential matches, studies have repeatedly shown that accuracy drops sharply when the input image is blurry, taken at an angle, or when the subject wears hats, glasses, or masks. Error rates are often higher for people of color, women, and younger individuals. In this instance, the suspect in the video wore a cap and mask, conditions that multiple independent evaluations have identified as especially problematic for current algorithms.
The Ars Technica report notes that Williams is Black, adding another layer to concerns about demographic disparities in facial recognition performance. Research conducted by the National Institute of Standards and Technology has found that many commercial systems exhibit significantly higher false positive rates for African American and Asian faces compared with Caucasian faces. These disparities become especially concerning when police treat a match as probable cause without additional verification.
Williams’ legal filing argues that the Palm Beach County Sheriff’s Office had clear policies requiring investigators to corroborate facial recognition results with traditional evidence. According to the suit, those policies were not followed. Instead, once the software identified Williams, officers appeared to build the remainder of their case around that single data point. The homeowner’s identification in the photo array may itself have been tainted, the complaint suggests, because she had already been shown Williams’ driver’s license photo linked to the facial recognition match. Defense attorneys frequently challenge such procedures as suggestive.
The case also raises questions about how agencies document and disclose their use of facial recognition to prosecutors and defense counsel. In many jurisdictions, police reports do not always mention that an arrest originated from an algorithmic match. This lack of transparency can prevent defense attorneys from effectively challenging the reliability of the underlying evidence. Williams’ lawyers claim that the sheriff’s office only acknowledged the role of facial recognition after they specifically requested information about the investigation’s origins.
Broader data on facial recognition errors provide context for Williams’ experience. A 2019 report by the American Civil Liberties Union found that the technology had wrongly identified 26 members of Congress when their photos were compared against a database of criminal mug shots. Several high-profile cases have involved mistaken arrests based on faulty matches, including the 2020 arrest of Robert Julian-Borchak Williams in Michigan. That man, who shares the same first and last name as the Florida plaintiff, spent 30 hours in jail after Detroit police relied on a facial recognition match that turned out to be incorrect. The Michigan case drew national attention and prompted several cities to impose moratoriums on police use of the technology.
Florida has taken a different approach. The state maintains one of the most extensive facial recognition databases in the country, with strong support from law enforcement leadership. The Florida Department of Law Enforcement has defended the technology as a valuable investigative aid that helps solve cases faster. Officials point to instances in which facial recognition has correctly identified suspects in serious crimes including murders and sexual assaults. Yet critics argue that the absence of uniform standards for verifying matches before making arrests creates predictable risks of wrongful detention.
The Palm Beach County Sheriff’s Office has not publicly commented on the specifics of the Williams lawsuit, citing pending litigation. In previous statements about its biometric programs, the agency has said it uses facial recognition only as one tool among many and requires officers to seek additional evidence before seeking warrants. Williams’ attorneys counter that those assurances appear to have been ignored in practice. They are seeking damages for false arrest, malicious prosecution, and emotional distress. The suit also asks the court to order changes in departmental policy to ensure that facial recognition matches receive meaningful human scrutiny and that exculpatory evidence is pursued with equal vigor.
Legal experts following the case say it could influence how courts evaluate qualified immunity claims in future facial recognition disputes. Police officers are often shielded from civil liability when they act in accordance with established procedures, even if those procedures later prove flawed. If Williams can show that officers deliberately disregarded clear contradictory evidence, that defense may weaken. The outcome may also affect procurement decisions by other agencies considering similar technology.
Beyond the courtroom, the incident adds to a growing body of evidence that facial recognition systems require stronger guardrails when deployed in criminal investigations. Some researchers have called for mandatory human review protocols that include side-by-side comparison of original video, candidate photos, and contextual information such as timestamps and location data. Others advocate for independent testing of any system before it is used to support arrests. A few municipalities have gone further, banning government use of facial recognition outright until accuracy and bias issues are resolved.
Williams has described his time in jail as humiliating and frightening. He lost wages, faced public embarrassment when his arrest appeared in local news, and continues to experience anxiety about future encounters with law enforcement. His attorneys argue that these harms were entirely avoidable if investigators had simply checked the employment records that were readily available. The fact that such records existed and were ignored, they say, demonstrates a troubling willingness to let software substitute for basic police work.
The lawsuit arrives at a moment when federal lawmakers are considering legislation that would regulate law enforcement use of biometric identification. Bills introduced in Congress would require agencies to disclose their use of facial recognition, establish minimum accuracy thresholds, and create mechanisms for individuals to challenge erroneous matches. Similar proposals have been debated at the state level, though Florida has not moved to restrict the technology.
For now, Richard Williams’ case serves as a concrete example of what can happen when enthusiasm for new investigative tools outpaces careful implementation. The package thefts that triggered the investigation remain unsolved. The actual perpetrator may never be identified. Meanwhile, an innocent man spent nearly a month in custody, his name entered into public records as a suspected thief, because a computer program saw a resemblance that human judgment should have questioned.
As more police departments adopt facial recognition, stories like this one are likely to multiply unless agencies adopt clearer protocols for verification. The technology itself is neither inherently good nor bad; its value depends entirely on how officers choose to use it. In this instance, according to the allegations, that choice produced a miscarriage of justice that three weeks of incarceration could not undo. The civil case will determine whether the Palm Beach County Sheriff’s Office bears legal responsibility for those decisions and whether future suspects will receive more thorough protection against similar errors.
WebProNews is an iEntry Publication