Angela Lipps lost nearly six months of her life, her home, her car, and her dog because a facial recognition system told Fargo police she was someone she wasn't.
Lipps, a grandmother living in Tennessee, was arrested and jailed after the Fargo Police Department's facial recognition technology flagged her as a suspect in a bank fraud investigation in the Fargo metro area. The problem: she was in Tennessee the entire time. She had nothing to do with the crime.
She spent months locked up across facilities in both Tennessee and North Dakota before records finally confirmed what she'd been saying all along. The charges were dismissed. But by that point, the damage was done.
The Cost of a False Match
Facial recognition systems work by comparing a photo against a database of faces, looking for statistical matches. These systems are faster than human investigators, but they also make mistakes that humans might not. Multiple studies, including a landmark 2019 NIST evaluation, have shown that many facial recognition algorithms have higher error rates for women, older adults, and people of color.
In Lipps' case, the AI's confidence in its match apparently carried enough weight that investigators moved forward with an arrest. What's missing from this story is any indication that Fargo police independently verified the match before acting on it. A basic records check showing Lipps' location in Tennessee could have prevented the entire ordeal.
No Algorithm Should Be Probable Cause
This case puts a sharp point on a question cities across the country are already wrestling with: should law enforcement be allowed to arrest people based primarily on an AI match?
San Francisco, Boston, and several other cities have banned or restricted police use of facial recognition. The European Union's AI Act classifies real-time facial recognition by law enforcement as "high risk" and imposes strict conditions on its use. North Dakota has no such restrictions.
The pattern here is familiar. An AI system produces an output. Humans treat that output as fact rather than as one data point that needs verification. Someone's life gets destroyed in the gap between algorithmic confidence and actual accuracy.
Lipps is now working to rebuild her life. She lost months of freedom and the basic stability of having a home and possessions, all because of a software match that was wrong. No word yet on whether she plans to pursue legal action against the Fargo Police Department, but cases like these have resulted in settlements elsewhere. Robert Williams received a settlement from Detroit after a similar wrongful arrest based on facial recognition in 2020.
For anyone building or deploying AI systems, this is the kind of outcome that should inform every design decision. The technology isn't the whole problem. The problem is deploying it without guardrails and treating its output as ground truth.