federal challenges to doj program reveal glitch: AI-generated court cases share ghost fingerprints with missing persons database
Researchers analyzing the DOJ's flagship crime-fighting AI have stumbled on a bizarre glitch: hundreds of "fictional" court documents used in the program's training data contain metadata timestamps and latent fingerprint partials that exactly match unsolved missing persons cases from three separate states. The algorithm, now facing federal challenges, appears to have been trained over a digital graveyard.