The face machine goes shopping

From 2012 to 2020, Rite Aid deployed facial-recognition technology in hundreds of stores. In its complaint, the Federal Trade Commission alleged the system generated thousands of false-positive matches. Employees followed customers, searched them, ordered them out, and sometimes contacted police based on matches that were wrong.

The FTC said the system was more likely to generate false positives in stores located in plurality-Black and Asian communities than in plurality-White communities. The complaint described low-quality enrollment images, a lack of meaningful accuracy testing, inadequate employee training, and failures to track what happened after alerts.

The software was not handed a microscope. It was handed authority, bad photographs, weak controls, and the public.

One face, one thousand alerts

The FTC complaint gives the bureaucracy a human scale. One image enrolled at a Bronx store generated more than 1,000 alerts in roughly two months—nearly five percent of all match alerts across the system during that period. That should have been a screaming alarm about the enrollment. Instead, the machinery continued to produce suspicion.

The FTC's proposed order would impose a five-year ban on using facial recognition for surveillance, along with deletion, notification, monitoring, and security requirements. The documented allegations remain a model of what happens when experimental identification becomes operational authority.

CRR remedy

High-impact surveillance needs independent accuracy testing, demographic performance analysis, strict image-quality rules, trained operators, recorded outcomes, and a rapid way for affected people to challenge errors. Better still, ask whether the system should exist in that setting at all.

The face machine requests reassignment to counting toothpaste, a task for which management already possessed barcodes.