Face-as-ID Goes Mainstream. Accuracy Hasn't.

Three separate announcements dropped this week confirming that facial recognition is no longer a pilot program—it's infrastructure. TSA launched a second facial recognition trial at Las Vegas's McCarran International Airport. Alaska Airlines rolled out automated facial ID checks at bag drop units in Seattle and Portland. Panasonic and JR East kicked off a proof-of-concept for walk-through facial ticket gates at Nagaoka Station on Japan's Joetsu Shinkansen. Your face, officially, is the new boarding pass.

Then came the other half of the news cycle. And it was considerably less reassuring.

A Very Busy Week for Your Face

Let's run through what actually happened, because the volume of announcements in a single week is itself the story.

At Las Vegas McCarran, FEDagent reported that TSA launched a 30-day proof of concept for automating identity verification at checkpoints—comparing a live facial scan against the photo on a traveler's ID document. Participation is voluntary. Travelers who opt out go through standard screening. The agency's Privacy Impact Assessment documents exactly what gets collected: real-time facial image, ID photo, document issuance and expiration dates, date of travel, document type, issuing organization, and year of birth. That's a fairly substantial data grab for a "voluntary" pilot.

Meanwhile, Alaska Airlines quietly started running facial ID checks at automated bag drop kiosks in Seattle and Portland. No press conference, no big launch event—just a new step in the check-in process. And in Japan, Panasonic Connect announced a trial at Nagaoka Station that goes a step further: walk-through ticket gates with synchronized visual and audio effects, no card tap required. JR East is framing this as part of its "Suica Renaissance" initiative—an evolution beyond the IC card system toward a fully biometric transit experience. It's genuinely impressive engineering. It's also a significant shift in how a country manages mass transit identity.

Three deployments, three different use cases, three different continents. All in the same week. The message from the industry is unmistakable: this is normal now. This article is part of a series — start with Eu Ai Act Facial Recognition 2026.

The Part Nobody's Advertising

Here's where it gets interesting—and uncomfortable.

While airports and rail operators were announcing smooth biometric futures, WIRED published internal records showing that Mobile Fortify—the facial recognition app currently deployed by ICE and CBP agents in towns and cities across the United States—was never designed to reliably identify people in the field. The Department of Homeland Security launched Mobile Fortify in spring 2025 to, in their words, "determine or verify" the identities of individuals stopped or detained during federal operations. DHS explicitly tied the rollout to an executive order directing a "total and efficient" crackdown on undocumented immigrants.

The problem? The app doesn't actually do what DHS says it does.

"Every manufacturer of this technology, every police department with a policy makes very clear that face recognition technology is not capable of providing a positive [identification]." — Internal records reviewed by WIRED

Read that again. The manufacturers of this technology—the people who built it and are paid to sell it—are on record saying it cannot provide a positive identification. And yet federal agents are using it in active enforcement operations, in the street, against real people, with real consequences. That's not a theoretical concern. That's a documented gap between what a system is marketed as doing and what it's technically capable of doing.

Separately, The Regulatory Review has been tracking legal critiques of TSA's facial systems, particularly around traveler rights and whether the "opt-out" framing holds up under scrutiny. Constitutional challenges to the process are accumulating. Courts are starting to ask questions that agencies don't have clean answers to.

Recognition vs. Comparison: The Distinction That Actually Matters

This is the part where most coverage loses the thread—and where anyone doing serious investigative or forensic work needs to plant their flag clearly. Previously in this series: Super Recognizers Facial Comparison Accuracy.

Facial recognition and facial comparison are not the same thing. Not remotely. Recognition means taking an unknown face and running it against a large database in real time, trying to find a match. It's a one-to-many search under variable, uncontrolled conditions—the person could be moving, the lighting could be terrible, the camera angle could be wrong. The error rates compound with every variable you can't control. This is what TSA is doing at checkpoints, what ICE is doing in the field with Mobile Fortify, and what makes civil liberties lawyers reach for their keyboards.

Facial comparison is fundamentally different. You start with two known images—a reference and a query—and you evaluate them against each other using geometric analysis: measuring the Euclidean distances between facial landmarks to determine whether they plausibly depict the same person. It's a one-to-one evaluation, conducted under documented conditions, with a defined methodology and a verifiable error rate. The science is older, better understood, and—critically—defensible in court when someone challenges your results.

The problem is that regulators, journalists, and even agency procurement officers routinely conflate these two things. When Mobile Fortify generates controversy and TSA facial scans end up in a constitutional challenge, every facial technology gets painted with the same brush. If you're doing controlled facial comparison work—the kind of methodical, documented face comparison that holds up under cross-examination—you need to be able to articulate that distinction quickly and clearly. Because right now, the reputational damage from mass deployment failures is going to splash.

The Operational Argument vs. The Forensic Argument

Look, nobody's saying mass facial deployment is purely incompetent. There's a reasonable operational argument for imperfect systems: even a tool with meaningful error rates can outperform an exhausted human agent checking documents manually at 3am on a busy travel weekend. Volume helps. Speed helps. Catching some fraud is better than catching none.

That argument makes sense if you're optimizing for throughput at a checkpoint. It falls apart completely if you're trying to use a result in court, in a deportation proceeding, or in any context where someone's rights, freedom, or safety are on the line. "Better than nothing" is not a forensic standard. It's not even close to one.

The New York Times noted this week that facial ID is accelerating at check-in counters across multiple carriers—framing it as convenience, as progress. And in the narrow sense of "is this faster than showing your passport," yes, it probably is. But the speed of the transaction is entirely disconnected from the reliability of the underlying identification. Panasonic's Shinkansen gates at Nagaoka Station look genuinely elegant—the synchronized visual and audio effects, the walk-through experience, the clean engineering. None of that tells you anything about what happens when the system makes a mistake at 120 mph between Tokyo and Niigata. Up next: Face As Id Went Mainstream This Week Accuracy Didn.

The credibility problem building around mass deployment isn't because the biometric science is fundamentally broken. It's because agencies skipped the step that makes results defensible: controlled conditions, documented methodology, verifiable error rates. They prioritized deployment speed over analytical rigor—and now the legal and reputational consequences are arriving on schedule.

The uncomfortable irony of this week's news is that the agencies generating the most controversy—DHS deploying Mobile Fortify without verified accuracy thresholds, TSA running trials that are generating constitutional challenges—have inadvertently made the case for exactly the kind of disciplined, methodology-first approach that serious investigators have always needed to practice. The controversy is a gift, in a way. It's making clients, judges, and internal counsel ask questions they should have been asking all along.

So here's the question worth sitting with: when a federal immigration agent uses a facial app that its own manufacturer says can't provide positive identification—and then acts on that result in an enforcement operation—what exactly is the evidentiary basis for what happened next? And if that case ends up in front of a judge, who's explaining the difference between what the technology can do and what the agency claimed it did?

Because right now, that explainer doesn't exist. And someone is going to have to write it.