On the morning of July 14, 2025, a team of U.S. Marshals arrived at Angela Lipps’s Tennessee home and arrested her at gunpoint while she was babysitting four young children. Lipps, a 50-year-old grandmother who had never flown on an airplane and had never set foot in North Dakota, found herself charged with eight felony counts of bank fraud in Fargo — a city she had never visited. The arrest warrant rested substantially on a facial recognition match generated by Clearview AI, a commercial system operating on a database of billions of photos scraped from the internet. No one from the Fargo Police Department called to question Angela before the warrant was issued. For 108 days, she sat in a Tennessee jail cell without bail, held as an accused fugitive awaiting extradition to face charges in a state she had never entered. When authorities finally reviewed her bank records in December, they revealed what any interview could have established months earlier: Lipps had been in Tennessee the entire time, her transactions placing her squarely at home while the actual fraud was occurring twelve hundred miles away. The charges were dismissed on Christmas Eve. The Fargo Police Department offered no apology and no explanation for why no investigator had spoken with her during five months of incarceration. The Fargo case is more than a cautionary tale about one flawed investigation. It exposes a growing institutional habit of treating AI outputs not as fallible signals requiring human verification, but as authoritative conclusions carrying near-prosecutorial weight. A Fargo detective reviewed Lipps’s social media and driver’s license, then concluded she matched the suspect based on “facial features, body type and hairstyle and color” — ratifying what the machine had decided rather than examining the evidence independently. The algorithm had become the accuser, and every