Retailers have been clear about the problem. Theft is rising, margins are tightening, and stores are adjusting. What is less visible is how those adjustments are changing the environment customers move through every day — and who absorbs the cost of that change most directly.
Major chains are expanding investments in AI-driven surveillance systems, including advanced camera networks, behavioral tracking software, and facial recognition tools. These systems are designed to identify patterns associated with theft, flag suspicious behavior, and assist with coordinated enforcement across multiple locations. The goal is not just prevention. It is prediction. That distinction matters, because predictive systems do not wait for behavior to occur — they interpret presence itself as data, assigning risk before any action is taken.
The racial stakes of that interpretation are documented and ongoing. The Innocence Project has confirmed at least seven cases of misidentification due to facial recognition technology, six of which involve Black people who were wrongfully accused. A 2025 study of facial recognition accuracy in law enforcement contexts found that error rates are consistently higher for women and Black individuals, with Black females most affected — a pattern that has persisted across multiple generations of the technology and multiple evaluations by independent researchers. When those systems are embedded in retail enforcement, they do not simply flag behavior. They distribute suspicion unevenly before a single interaction occurs.
For some shoppers, these systems remain invisible. For others, they are felt directly — through increased scrutiny, locked merchandise, altered store layouts designed to channel movement, or the particular experience of being watched in ways that others nearby are not. This is not a secondary concern to be addressed after the efficiency question. It is the central question. The communities most likely to be misread by these systems are the same ones that have historically faced profiling, presumption of guilt, and over-policing in commercial spaces. Technology does not resolve that history. Deployed without accountability, it encodes it into automated infrastructure that operates at scale and without the friction of human discretion.
What is also underexamined is the data trail these systems generate beyond their stated purpose. The camera networks, behavioral databases, and predictive models built for loss prevention create persistent records of customer presence, movement patterns, and interaction histories. That data does not disappear when someone leaves the store. Depending on how it is stored, shared, or sold, it can feed systems far outside the retail context — contributing to the kind of ambient surveillance infrastructure that we examined in our coverage of algorithmic tools and community-level data extraction. Shoppers are moving through spaces that generate detailed records of their behavior without meaningful disclosure or consent, and without any clear legal framework governing what happens to that information next.
Retailers are responding to real financial pressure. But enforcement infrastructure rarely stays contained to its original purpose, and the communities with the least institutional power to push back against misidentification, data misuse, or discriminatory profiling are the ones most likely to encounter these systems at their most consequential. The shift from loss prevention to predictive surveillance is not a technical upgrade. It is a structural one — and the structural questions deserve the same attention as the operational ones.