Facial Recognition Isn't on Trial. Your Explanation Is.

Illinois is moving to ban police from using facial recognition. The TSA, meanwhile, is scanning faces at more than 250 airports across the country. Same core technology. Opposite policy directions. And right in the middle of this mess: investigators, prosecutors, and private analysts who use facial comparison every day and are now being asked — sometimes by their own department heads — to explain why their tool isn't the surveillance technology making headlines.

Two Policies, One Technology

Illinois House Bill 5521 recently gained two new co-sponsors and is gaining momentum despite failing to clear committee in March 2026, according to CWB Chicago. The bill would bar law enforcement from using facial recognition as an investigative tool entirely — no database searches, no lead generation, no matching a surveillance still to a known offender. Illinois would join a growing list: as of early 2025, 15 states had enacted some form of law restricting facial recognition in policing, with more than 16 cities implementing outright bans, per State of Surveillance. San Francisco drew the national attention when it moved first in 2019. Now it has company.

Federal agencies aren't reading the same memo. The TSA explicitly frames its deployment as "facial comparison technology" — not recognition, notably — and states in its official fact sheet that the technology "will not be used for surveillance or any law enforcement purpose." That's a careful distinction. The system automates what a TSA agent was already doing manually: checking your face against your ID. No database search. No watch list comparison. No sending your face somewhere else.

The problem is that the public — and plenty of lawmakers — don't see the difference. They hear "facial recognition" and they think of the worst version of it. And honestly? That's partly because the worst version of it actually happened.

The Case That Changed Everything

Robert Williams. Remember that name. In 2020, this 45-year-old Black man was arrested in Detroit after an investigator relied on a facial recognition match as the primary basis for an arrest warrant. The system had misidentified him. He spent time in custody for a crime he had nothing to do with. His case didn't just become a cautionary tale — it became the legislative fuel behind nearly every state ban that followed, as documented by the American Bar Association. This article is part of a series — start with China Made Creating A Deepfake The Crime Not Sharing It U S .

Here's the thing though: the failure in the Williams case wasn't the technology running unsupervised. It was a detective treating a probabilistic match as a definitive ID. That's a methodology problem, not a technology problem. The distinction matters enormously — and it's exactly what critics of a blanket ban keep pointing out.

"Facial recognition is one of the most important investigative tools to come along in policing in 50 years." — Retired Riverside Police Chief Tom Weitzel, as reported by Chicago Today

Weitzel's position is one that most working investigators share privately even if they've learned to express it more carefully in public. The tool works. Used correctly — as a lead generator, not a verdict — it solves cases that would otherwise go cold. Homicides. Missing persons. Child exploitation. The Illinois bill's supporters have struggled to produce local examples of the technology being misused in the state, which is part of why it stalled in committee the first time around.

The Distinction That's Getting Lost

Context is everything here, and the regulatory divergence reflects how completely context is being ignored in public debate. Washington State got this partly right: its law prohibits law enforcement from using facial recognition as the sole basis for establishing probable cause. That's a guardrail, not a ban. It says: use the tool, but don't let it be the only thing standing between a match and an arrest warrant. New Jersey, Maryland, and Montana have all moved in a similar direction, adding disclosure requirements so that defendants are notified when facial recognition played a role in their case — an approach the Biometric Update has covered as a meaningful middle ground.

The TSA's framing, meanwhile, is instructive for anyone who works in facial comparison professionally. Notice they don't call it "facial recognition." They call it "facial comparison technology." That's not marketing spin — it's a functional description. The system compares your face to your government-issued ID credential. One known image, one live capture, one match. No database. No population-level sweep. No continuous monitoring of a crowd.

That's essentially what solo investigators and small forensic teams do when they use controlled facial comparison in casework. You have a subject photo. You have a reference image. You're asking: is this the same person? It's methodologically closer to fingerprint analysis than to surveillance. But try explaining that to a judge who just read three headlines about facial recognition bans, or to a client whose last exposure to this technology was a documentary about authoritarian governments scanning protesters. Previously in this series: The Courtroom Question Youre Not Ready For Prove This Video .

The Explanation Gap Is the Real Risk

Look, nobody working in facial comparison thinks their tool is infallible. That's not the argument. The argument is that a carefully controlled comparison — case-specific images, known subjects, documented methodology, human analyst review — is categorically different from running an unknown face through a database of millions and treating the top result as a suspect. One is forensic analysis. The other, deployed badly, is what put Robert Williams in handcuffs.

The investigators who will come out of this regulatory moment in the best position aren't necessarily the ones using the most advanced tools. They're the ones who can explain — clearly, specifically, without jargon — exactly what their process does and doesn't involve. What images were used. What the system returned. What human review followed. What corroboration existed before any action was taken.

Tools like CaraComp are built around exactly this kind of transparent, case-specific facial comparison workflow — designed to produce court-ready documentation that shows not just the result, but the methodology behind it. In an environment where "facial recognition" has become a politically loaded term, the ability to show a judge or a client a clear audit trail of your process isn't a nice-to-have. It's your professional defense.

Opponents of these bans are also right that transparency requirements are the more sensible fix, as The Regulatory Review has noted in its coverage of TSA's expansion. The Algorithmic Justice League found that most airport passengers receive little or no meaningful notice about their ability to opt out of face scans, with signage using vague phrases like "biometric identity technology" rather than plainly saying "facial recognition." That's a consent problem. Fix the consent problem. Don't throw out the investigative capability that has solved violent crimes and located missing children.

The Question on the Table

Here's what I want to know from the people doing this work every day. As Illinois advances its ban and federal agencies push ahead with biometric ID at airports and immigration checkpoints, the "facial recognition" category has become too broad to be useful in public conversation. It's being applied equally to a detective running a crime scene still through a database and a border agent confirming a passport holder's identity and a private investigator comparing two photos in a civil case. These are not the same thing.

So the question isn't whether facial recognition should be regulated. Of course it should. The question is whether you have a clear, specific, jargon-free explanation ready for when a judge, a client, or your department head asks you to justify your methodology — and whether that explanation includes documented proof of exactly how the comparison was conducted and what human review followed.

Because right now, the lawmakers drafting these bans aren't making a distinction between Robert Williams' case and yours. If you don't make that distinction for them, someone else will make it for you — and probably not in your favor.

How are you drawing that line in your own work — and who are you drawing it for?