Mobile Fortify: US Agencies Are Guessing at Your Face

Imagine you are standing at a border, or simply walking down the street, and an officer points a smartphone at you. A couple of seconds later, an algorithm hands down its verdict. It sounds like a cyberpunk scene, but for the US it is the daily reality of the Mobile Fortify app. The catch is that the app is technically incapable of confirming identity with adequate accuracy. It was simply not built for that. That did not stop ICE and Customs and Border Protection officers from using it more than 100,000 times to check both migrants and US citizens.

The story of Mobile Fortify is a textbook example of how the urge to adopt innovation at any cost turns serious government institutions into characters in an absurdist comedy. The system was originally conceived as an auxiliary tool, a kind of digital notepad with access to databases. At some point it became the main hammer used on every nail.

To legalize the practice, the Department of Homeland Security had to resort to a small trick: it simply ignored its own privacy impact assessment rules. When rules get in the way of progress, bureaucrats do not change the progress, they scrap the rules.

Why does this matter right now? We are at the peak of the AI-tool hype, and government bodies around the world are trying to automate everything within reach. But when it comes to facial recognition, the cost of a mistake is not just a badly chosen recommendation in a news feed. It means wrongful detentions, interrogations and ruined lives. If a system returns only a probabilistic answer and an officer treats it as the ultimate truth, the technology is more dangerous than no technology at all.

The most ironic part is that the agency itself acknowledged that Mobile Fortify did not pass tests for compliance with identification standards. It is as if police used tea-leaf reading instead of a breathalyzer, yet filled out the paperwork with a straight face. The system often confuses people with similar facial features, which, when working with migrants, turns into a real lottery. Meanwhile, scan data is stored in databases that are later used to train other, even more invasive algorithms.

We are seeing a dangerous precedent in which a "kludge" becomes the de facto standard. Facial recognition already draws heavy criticism for bias and errors affecting ethnic minorities. Using software that is knowingly faulty or unsuitable only adds fuel to the fire. Instead of building a transparent oversight system, the authorities chose the path of least resistance, where interface convenience outranks civil rights.

It is a reminder that behind any AI "magic" there is specific code and specific people who approve that code. If, at the approval stage, common sense is replaced by the wish to tick a box in an innovation report, the result is a digital dystopia at the scale of an entire country. And while human rights advocates try to challenge the use of Mobile Fortify, the counter keeps ticking, logging more and more faces in a system that officially "does not work as it should."

The bottom line: if even US agencies are ready to use questionable software while bypassing their own rules, how far can we trust any other government AI system without an independent audit?