Facial Recognition's Real Reckoning: Courts Want a Paper Trail

At least twelve people in the U.S. have now been wrongfully arrested because of facial recognition. A twenty-twenty-five Washington Post investigation found that in six of those eight documented cases at the time, police never even checked the subject's alibi. They skipped the most basic step in detective work.

That pattern matters to anyone who works in

That pattern matters to anyone who works in investigations, biometrics, or digital identity. It matters if you use these tools. It matters if you're ever on the other side of them. The story centers on Angela Lipps, the eighth documented wrongful arrest tied to facial recognition in the U.S. She's white — which breaks the assumption that these failures only affect Black Americans. Most previous victims were Black, and algorithmic bias played a role. But Lipps' case shows something bigger: the system itself — the human decisions around the algorithm — is broken. So the question threading through all of this is — what happens when courts stop asking "Was the match accurate?" and start asking "Show me your paperwork"?

Start with Fargo, North Dakota. After the Lipps arrest, Fargo's police chief, Dave Zibolski, pulled his department off the facial recognition system run by neighboring West Fargo. His reason was blunt: "We don't know how it's run or how it's overseen." Instead, Fargo now routes searches through the state's certified center — a facility with trained analysts and documented procedures. That's not a department dropping the technology. That's a department demanding a paper trail before it trusts the results.

And Fargo isn't alone. In Detroit, the settlement in the Robert Williams wrongful arrest case produced what advocates call the nation's strongest police department policies constraining facial recognition. Those policies require documentation at every step — comparison logs, match confidence thresholds, independent corroboration before any arrest. The shift is clear. Judges and prosecutors are moving toward a standard: if you can't show how you got your match, your evidence doesn't come in.

Meanwhile, the Washington Post findings paint a damning picture of what happens without those guardrails. In two of the wrongful arrest cases, officers ignored contradictory evidence that pointed away from the subject. In five, they failed to collect key evidence entirely. The algorithm gave a suggestion, and officers treated it like a conclusion. That's not a technology problem. That's a governance collapse.

Zoom out past the U

Now zoom out past the U.S. On 03-17-2026, Brazil began enforcing its Digital Statute for Children and Adolescents — the Digital E.C.A. The law requires auditable biometric age assurance, meaning companies must prove how they verified a user's age and log that proof. Penalties are severe — fines up to fifty million Brazilian reais, or ten percent of a company's revenue. Brazil's data protection authority can also shut a service down entirely. The guidelines specifically flag facial biometrics as high-risk because of surveillance exposure and algorithmic bias. Definitive rules arrive in August twenty-twenty-six after public consultations.

The U.K. is pushing in the same direction. Its Online Safety Act now extends age verification requirements to platforms like Reddit, Discord, Spotify, and X. The goal isn't to mandate one method. It's to build an interoperable ecosystem — documents, biometrics, encrypted tokens — all auditable, all standardized across borders. So whether you're an investigator in Detroit or a compliance officer in São Paulo, the expectation converges on the same point: show your work.

Law enforcement pushes back, of course. Documentation adds friction. Audit trails slow things down when speed matters. But the evidence cuts against that argument — hard.

The real danger was never slow investigations. It was fast, wrong ones — arrests that destroyed lives and then fell apart in court.