When Kimberlee Williams accompanied her daughter to an Oklahoma military base in 2021, security officers learned that she was wanted for bank fraud in Maryland. Facial-recognition software had concluded Williams was the person responsible for impersonating individuals in Maryland and withdrawing thousands of dollars from their bank accounts. So she was taken into custody and held in an Oklahoma jail for 23 days — and then spent months more behind bars in Maryland. Download the SAN app today to stay up-to-date with Unbiased. Straight Facts™. Point phone camera here But Williams had never been to Maryland before her arrest. And she wasn’t the person the facial-recognition program thought she was. Now Williams is demanding an apology from police, as well as reforms to prevent others from being wrongly identified in criminal cases. Williams’ ordeal illustrates the flaws in commonly used facial-recognition technology — and, according to the American Civil Liberties Union, the mistaken belief by many law enforcement agencies that a match in the technology indisputably verifies a person is a wanted criminal. “But this kind of ‘verification’ is worthless,” the ACLU and its Maryland chapter said in letters sent Tuesday to three Maryland police departments on Williams’ behalf. “It merely confirms that the technology did what it is designed to do: find someone who looks similar to the suspect, but is often innocent.” Mistaken identity Williams was riding along with her daughter, a DoorDash driver, for a delivery to Fort Sill in Lawton, Oklahoma, when security officers performed a routine ID check before allowing the women onto the base. The check revealed a warrant for Williams’ arrest, based on an image of a suspect that a bank investigator had uploaded to an electronic mailing list tool used by law enforcement and private investigators. An individual who saw the image