Key Takeaways
- Kimberlee Williams, an Oklahoma resident, was wrongfully arrested and jailed for six months after Maryland police relied on an erroneous facial‑recognition match.
- The detective who sought the arrest warrant concealed that the identification came solely from an unverified facial‑recognition search and instead presented a misleading visual comparison to the magistrate.
- Charges were ultimately dropped in Montgomery, Prince George’s, and Anne Arundel counties, but Williams spent a total of 180 days in custody away from her family, job, and home.
- The ACLU and ACLU of Maryland have demanded a public apology, an independent investigation of the procedural failures, and policy reforms that ban reliance on outside facial‑recognition leads and prohibit arrests based solely on algorithmic matches.
- Williams is the 14th publicly known case of wrongful arrest in the U.S. stemming from faulty facial‑recognition technology, highlighting a systemic risk that disproportionately affects people of color, women, older adults, and youths.
Background of the Incident
On June 23, 2021, Kimberlee Williams was accompanying her daughter, a DoorDash driver, during a delivery to a military base in Lawton, Oklahoma. Base security performed a routine ID check and discovered an outstanding arrest warrant issued by Maryland authorities. Williams was detained by local Oklahoma police and held for 23 days before a Maryland officer transported her to Montgomery County, where she was jailed for over three months while she protested her innocence. The warrant alleged that Williams had impersonated bank customers and withdrawn thousands of dollars from several Maryland bank branches—a crime she had never committed and could not have committed, given she had never set foot in Maryland prior to her arrest.
How the Facial‑Recognition Error Occurred
The investigation leading to the warrant began when a bank investigator forwarded a still‑image of the unknown suspect to a national listserv called Crimedex. An unidentified subscriber ran the image through facial‑recognition software and returned Williams’s name and photograph as a purported match. The bank investigator then memoed a Montgomery County detective, stating that “facial recognition software” had flagged Williams as the suspect. Acting on this tip alone, the detective applied for an arrest warrant without seeking any independent corroboration—such as surveillance footage, transaction records, or witness testimony—to verify that Williams was actually the person seen in the banks.
Misrepresentation to the Court
In the warrant application, the detective deliberately obscured the true basis of the identification. Instead of disclosing that the lead came from an unverified facial‑recognition search conducted by an unknown entity, he claimed that Williams had been “identified” as the suspect and that he had personally confirmed the match by visually comparing a photo of the suspect with an older photo of Williams. Because facial‑recognition systems often produce false matches when the suspect merely resembles an innocent individual, this visual comparison was meaningless; it merely reinforced the algorithm’s error. Had the detective been transparent about the unreliable nature of the facial‑recognition result—or conducted a basic investigative check—the magistrate judge would likely have denied the warrant, preventing Williams’s wrongful arrest.
Legal Proceedings and Release
Montgomery County prosecutors dismissed the charges against Williams in October 2021 after recognizing the lack of credible evidence. However, the same faulty facial‑recognition match had spawned pending charges in Prince George’s and Anne Arundel counties, where additional bank branches had reported similar fraud. Williams was transferred from Montgomery County’s jail to Prince George’s County’s facility, where she remained incarcerated for another two months while contesting the allegations. All charges across the three counties were finally dropped in December 2021. In total, Williams endured six months of confinement—time she lost with her children, her employment, and her Oklahoma home—despite never having visited Maryland before her arrest.
ACLU Involvement and Demands
The American Civil Liberties Union (ACLU) and its Maryland affiliate intervened on Williams’s behalf, sending letters to the police departments of Montgomery, Prince George’s, and Anne Arundel counties. The letters demand:
- A thorough, independent investigation into the procedural failures that led to her wrongful arrest.
- A public apology acknowledging the harm caused by reliance on flawed facial‑recognition technology.
- Policy reforms that prohibit police from acting on facial‑recognition leads generated by outside entities without independent verification.
- A ban on arrests based solely on a facial‑recognition match followed by a superficial human comparison, which is inherently tainted when the algorithm falsely matches an innocent look‑alike.
The ACLU emphasizes that these reforms are necessary not only to redress Williams’s specific injustice but also to prevent similar abuses across Maryland and the nation.
Broader Context of Facial‑Recognition Misuse
Williams’s case is the 14th publicly documented instance of a wrongful arrest in the United States stemming from erroneous facial‑recognition results. Similar incidents have been reported in police departments across Detroit, Michigan; New Orleans, Louisiana; Las Vegas, Nevada; and other jurisdictions in Michigan, Missouri, New Jersey, North Dakota, Florida, and Arizona. Studies consistently show that facial‑recognition algorithms exhibit higher false‑match rates for people of color, women, older adults, and youths—demographics that Williams embodies as a Black woman. In 2024, the ACLU secured a landmark settlement with the City of Detroit after the wrongful arrest of Robert Williams, underscoring a growing recognition of the technology’s dangers.
Legislative and Policy Gaps
David Rocah, senior staff attorney at the ACLU of Maryland, noted that Maryland’s 2024 legislation governing facial‑recognition use lacks clear directives on what additional investigative steps must follow a facial‑recognition lead. Without explicit requirements for corroborating evidence, police departments are left to rely on the algorithm’s output alone, creating a pathway for mistakes like Williams’s arrest. Rocah warned that until the state mandates rigorous verification—such as obtaining independent eyewitness accounts, reviewing transaction logs, or confirming alibis—similar injustices will persist.
Conclusion and Call to Action
Kimberlee Williams’s ordeal serves as a stark reminder that facial‑recognition technology, when used without proper safeguards, can devastate lives. Six months of unjust incarceration cannot be reclaimed, but accountability, transparency, and substantive policy change can mitigate future harm. The ACLU’s demands for an apology, independent investigation, and enforceable reforms represent a necessary step toward ensuring that law‑enforcement agencies treat facial‑recognition leads as investigative tips—not conclusive proof—and that the rights of individuals are protected against the fallibility of automated systems. As more jurisdictions grapple with the implications of this technology, Williams’s case underscores the urgent need for nationwide standards that prioritize accuracy, equity, and justice over convenience.