Can facial recognition get an innocent person arrested? Yes — and it already has, multiple times, with documented victims serving real jail time before anyone noticed the algorithm was wrong. Now, in 2026, New Jersey’s top court has had enough. The court ordered police departments to disclose how they’re using facial recognition technology — a ruling that cuts straight through the fog of secrecy law enforcement has been hiding behind for years.
This isn’t a minor procedural tweak. This is a court telling cops: show your work, or stop doing it.
How Facial Recognition Has Already Sent Innocent People to Jail
The wrongful arrest cases tied to facial recognition aren’t hypothetical. Robert Williams in Michigan. Randal Reid in Georgia. Nijeer Parks in New Jersey. All Black men. All arrested based on facial recognition matches that turned out to be flat wrong. All locked up — one for nearly a week — before investigators bothered to look past the algorithm.
Facial recognition wrongful arrests disproportionately affect people with darker skin tones because most systems were trained on datasets that skewed heavily white and male. That’s not a glitch. That’s a feature of how these tools were built and then rushed into police departments with zero standardized oversight.
The technology itself isn’t the whole problem. The problem is cops treating a probability score like a conviction. A facial recognition match is not proof. It is a lead. Law enforcement agencies that failed to understand that distinction put innocent people in handcuffs.
And here’s the part that should make you furious: in most states, there was no requirement to even tell a defendant that facial recognition was used to identify them. Defense attorneys couldn’t challenge the method. Juries didn’t know it existed. The algorithm was just quietly baked into the prosecution’s case and presented as human detective work.
What New Jersey’s Court Order Actually Requires in 2026
The New Jersey Supreme Court’s ruling mandates that prosecutors disclose when facial recognition technology was used in building a case against a defendant. That disclosure requirement is a direct response to the opacity that enabled wrongful arrests in the first place.
Defense attorneys now have the legal ground to scrutinize the technology — which vendor supplied it, what accuracy rates the system claims, what the match confidence score actually was, and whether the human detective who reviewed the result received any training to interpret it critically. That’s a meaningful shift.
It also forces a conversation that police departments have been aggressively avoiding. Several NJ agencies were using commercial facial recognition tools — including some sourced from platforms that had rapidly scaled AI infrastructure through major enterprise partnerships — without any public procurement process or council approval. The public had no idea what tools were being run on their tax dollars, against their faces.
The ruling doesn’t ban the technology. That’s the compromise. But mandatory disclosure is the foundation everything else has to be built on. You cannot regulate what you cannot see.
Why Police Secrecy Around This Technology Was Always Indefensible
Law enforcement agencies have spent years arguing that revealing their facial recognition use would compromise investigations. That argument doesn’t hold up. Every other forensic method — DNA analysis, fingerprint matching, ballistics — is disclosed to defendants and subject to expert challenge. Facial recognition is forensic evidence. Treating it like a trade secret is a constitutional problem, not just a policy preference.
The real reason for the secrecy is simpler: these tools have embarrassing error rates, especially on non-white faces, and police departments didn’t want that scrutinized in court. The technology was convenient. It narrowed suspect pools fast. And as long as nobody knew it was being used, nobody could challenge it.
This connects to something broader happening across industries right now — the question of where AI accountability actually lives when automated systems affect real human outcomes. AI is reshaping decisions across dozens of sectors, from hiring to healthcare. In every one of those contexts, the same question applies: who is responsible when the algorithm is wrong? In policing, the cost of that question going unanswered is a person sitting in a cell for something they didn’t do.
The contrarian take you’ll hear from law enforcement advocates is that facial recognition, even with its flaws, solves crimes that would otherwise go cold. That’s probably true in some cases. But a risk-benefit calculation that treats wrongful imprisonment as an acceptable cost of doing business is not a technology argument — it’s a values argument. And it’s one that deserves to be made in public, not buried in procurement contracts.
Risk management frameworks in other industries — even construction — require systematic identification and disclosure of tools that could cause harm. The idea that policing gets a pass on this standard is a double standard that courts are finally starting to correct.
Where This Leaves Defendants, Departments, and the Rest of Us
The tension the New Jersey ruling opened up is this: disclosure without regulation is half a fix. Knowing that facial recognition was used is important. But if there’s no minimum accuracy standard, no training requirement for reviewing officers, and no limit on which vendors police can contract with, defendants are still fighting uphill.
Facial recognition wrongful arrests will not stop because of one court order. The wrongful arrest cases already documented represent a small fraction of the actual problem — most defendants don’t have the resources to discover how they were identified, let alone challenge it.
What the New Jersey ruling does is crack the door. It establishes that defendants have a right to know. Every state that hasn’t done this yet is operating a criminal justice system with a hidden variable that can end a person’s life before anyone notices the error.
The question at the start was whether facial recognition can get an innocent person arrested. It can. It has. It will again — until the gap between what this technology promises and what it actually delivers is treated as a legal problem, not just a technical one. New Jersey just made that argument in court and won. Now the rest of the country has to decide if it’s watching.