Artificial intelligence, at least according to its most prominent backers, is going to revolutionize every industry in the country—including the people tasked with protecting retailers from shoplifters. In a special edition of the trade publication Loss Prevention Magazine (LPM), which bills itself as the premier periodical for “asset protection professionals,” AI is presented as an exciting opportunity for the retail industry.
Sandwiched between jargon-laced articles about “Promoting a Data-Driven Culture” and how AI can help assess “voice stress to pinpoint possible signs of deception” among detained shoplifters, LPM assures readers that enhancing the local Whole Foods or REI with state-of-the-art surveillance tech is a “no-brainer.” After all, LPM has argued in the past that retail theft is on the rise across the country.
Such advances in AI would supposedly augment and speed up key tasks in retail security, like analyzing CCTV footage. “As opposed to the laborious process of watching hours of video, AI monitoring, for example, helps prevent theft at self-checkout kiosks by using cameras, sensors, and machine learning to analyze data and detect suspicious activity,” says one feature in LPM.
Though the chatter of a business-to-business magazine might seem benign, the pages of LPM reveal a candid portrait of how retail owners and management examine the social and economic landscape of the customers they serve and the workers they employ. For example, consider an LPM white paper on worker safety in retail: In the report, 1 in 4 workers are considering quitting their jobs not because of work conditions or stagnant wages, but because of the threat of amorphous, The Warriors-style gang members besieging Walmart.
Such developments portend a potential troubling future for retail employees, transforming their workplace—which employs more than a quarter of all U.S. workers—into an AI-powered panopticon. In this brave new world of worker surveillance, innocuous behavior will be criminalized, safety will be weaponized, and the ability to exercise one’s legally protected right to organize a union will be endangered.
WORKPLACE SURVEILLANCE, OF COURSE, is nothing new. At the turn of the 20th century, the hot new trend for manufacturing—similar in size then to retail today—was called “Taylorism.” This theory of employee supervision, pioneered by Frederick Taylor and also called “scientific management,” sought to maximize labor productivity as much as possible through elaborate tracking of workers’ time and movements. Managers could use stopwatches, time sheets, and other tools to ensure that not a single second of the working day would be “wasted.”
As the writer Robert Kanigel observed in his biography of Taylor, workers experienced even this rudimentary surveillance as tyrannical. A 1911 investigation by a Special House Committee concluded that the circumstances of Taylor’s methods were “the same as a slave driver’s whip on the negro, as it keeps [the worker] in a constant state of agitation.”
Taylor retorted that his system actually made workers more content in their labor. “I wish to state that until this last year, during the 30 years that scientific management has been gradually developed … there has never been a single strike of employees working under scientific management,” he told the Special Committee. Whether his system was actually just very good at controlling resentful people seems not to have occurred to him.
In this brave new world of worker surveillance, innocuous behavior will be criminalized and safety will be weaponized.
When workers did become recalcitrant, surveillance remained a key form of control. The Pinkertons, a notorious union-busting private security firm, used spying heavily to keep workers down in the late 19th and early 20th centuries. Their motto, “We Never Sleep,” often felt literal: Pinkerton spies were known not only to assault striking workers but also to infiltrate clandestine union meetings and collect names of all participants. (Incidentally, LPM praised the rebranded Pinkerton company in 2024. “Pinkerton stands at the forefront of retail security—vigilantly protecting what matters most in an industry where risks are as diverse as the products on the shelves,” an article from 2024 reads.)
Today, bosses don’t need armed thugs to enforce productivity rules; digital surveillance can do the job. Take the case of British banking conglomerate Barclays, which in 2020 was caught and fined for using heat and motion sensors to assess the amount of time its workers were spending at their desks.
As the academics Antonio Aloisi and Valerio De Stefano point out in their book Your Boss Is an Algorithm, we are living in a dystopia of employee monitoring. From ActivTrak, which “inspects the programs used and tells bosses if an employee is unfocussed, spending time on social media,” to Hubstaff and Sneek, which can “routinely take snapshots of employees through their webcams every five minutes or so to generate a timecard,” there is now a smorgasbord of Orwellian technology for managers.
Yet surveillance for loss prevention—which as an industry is quietly projected to double in value by 2030—is arguably even more pernicious due to its dubious motivations and the high likelihood that the data it gathers will be used for other purposes, particularly union busting. That this transformation in workplace surveillance coincided with sensationalized stories of lawless shoplifting sprees feels convenient for those seeking to profit from such consternations.
In the first place, it’s not clear that there is even a problem here. Following the post-pandemic surge in violent crime, lobbying groups like the National Retail Federation (NRF) leveraged the resulting anxiety, asserting that in 2021, there was supposedly $94.5 billion worth of inventory pilfered by organized crime rings preying on vulnerable retailers.
In reality, while retail crime did see an increase within some major American cities, overall there was no evidence for a national surge in shoplifting during this moral panic—especially not “organized retail theft.” The NRF later retracted its $94.5 billion claim and removed evidence that a 2020 study by the organization had found theft only cost retailers $720,000 for every $1 billion in sales.
Shoplifting rates do appear to be higher compared to pre-pandemic levels in places like New York, L.A., and Chicago. But that may have more to do with outside factors, like the cost-of-living crisis, and more importantly, there is reason to doubt AI surveillance will help at all.
SINCE THE SELF-CHECKOUT LANE became ubiquitous during the COVID-19 pandemic, workers in understaffed stores are now being deputized to maintain order. That’s where AI-powered technologies like Everseen come in. Everseen—which purports to use AI to flag insidious shoplifting behavior at self-checkouts and registers—has been deployed at retailers like Walmart for some time now. Anyone familiar with AI would not be surprised that Walmart employees have found this technology obnoxious and unreliable.
“You can imagine that adding a system that detects or stops shopper behavior that it deems to be ‘suspicious’ might still require some action from the frontline worker, whether that’s figuring out whether the behavior is actually suspicious, or dealing with a frustrated customer who’s been flagged for something,” Pegah Moradi, a Ph.D. candidate in information science at Cornell University, explained.
“If these flags are collected and easily provided to management, you can imagine that a worker might be under greater scrutiny during possible theft situations,” Moradi, who studies automation in the workplace, continued. “‘Why didn’t you do this?’, ‘Why did you do that?’ The power of the worker’s discretion is eroded when management trusts the AI signals more than the workers.”
Such an unreliable and potentially haphazard rollout of AI in the retail sector hasn’t stopped firms from buckling down. The U.S. Chamber of Commerce released a report praising the “Automated Worker Surveillance and Management” as a tool that could be instrumental in “Preventing Workplace Violence and Enhancing Safety and Security.” Meanwhile, Everseen filed a legal complaint in 2021 against Walmart, claiming that the superstore had stolen the firm’s trade secrets to develop “a Walmart copy of Everseen’s [Checkout Process Intelligence] technology.”
The AI surveillance hype contrasts with little attention paid to proven anti-shoplifting tools. As Amanda Mull writes at The Atlantic, understaffing, self-checkout machines, and poor logistics are all common in retailers, and all are associated with more theft. Progress could be made without the latest Big Tech gadget.
Meanwhile, stores like Walmart seem a lot more excited about using surveillance tools on their own employees than on criminals. A 2024 study of Walmart’s warehouse workers found that 45 percent felt “they always/most of the time have a sense of being monitored or watched on the job,” and that 58 percent of those polled “said the level of monitoring at Walmart is higher than at their previous job.”
Though such surveillance is ostensibly to protect the security of both workers and customers, it grants employers access to huge swaths of data that they can use for a variety of purposes. If stores monitor the traffic of customers as they move through the aisles, they can also track employees. It’s what information scientists Solon Barocas and Karen Levy have called “refractive surveillance,” wherein data collected from one group can be used to control another.
“This technology has been adopted and applied at such a rapid pace that the law has not caught up.”
“With [loss prevention] technology, they’re able to look at things like social media and past instances of theft or crime … and then create a different risk score of high, medium, and low for these different stores, and then allocate their loss prevention resources accordingly,” Madison Van Oort, an independent scholar, told the Prospect. Her research culminated in Worn Out, a book on retail surveillance and its consequences for workers.
“And so you can probably start to surmise, this is really similar to like predictive policing, right? It takes these stores that already have probably high theft or even instances of organizing, and then would continually allocate more research, more resources to those stores.”
Should workers at a retailer attempt to use social media to organize online or elevate their campaigns, Van Oort explained, these stores could be flagged to create higher risk scores for certain locations, and hence create yet another obstacle for union organizers. Without the ability to interact with each other without fear of surveillance, workers are unable to communicate with each other safely.
That’s how one Walmart employee, TaNeka Hightower, described her experience as a veteran Walmart employee and as someone who was a guinea pig for Walmart’s plan to have workers wear body cameras. Hightower has done everything from delivering groceries to working at her Nashville location’s vision center. And the way she tells it, the amount of monitoring at the store is “very detrimental to organizing co-workers.”
“We’re always under surveillance,” Hightower told me. “People have been fired for eating, drinking, or taking too long a break. We couldn’t talk to each other because even the pickup area where deliveries are received, the freezer, and the refrigerator are all [watched by cameras].”
According to Hightower, who has been with Walmart for seven years, this level of surveillance mixes with a punitive culture wherein threats about receiving disciplinary actions can be used to deny workers bonuses or cost them their jobs. “My file says I’m a troublemaker because they want people who are sheep, and I’m a goat,” Hightower said.
IF THIS ALL SOUNDS DISTURBINGLY LIKE how police behave, that might not be a coincidence. From the use of facial recognition software to the use of employee body cameras—which, in TaNeka’s case, appear to have been produced by Axon, the company that manufactures similar cameras for police and invented the taser—law enforcement tech is steadily encroaching into the retail sector.
At NRF Protect, the country’s premier loss prevention conference and trade show, retailers are encouraged to build lasting relationships with law enforcement by taking their local police chief out to lunch; customers and employees alike are painted as potential “bad guys” who need to be brought to justice; and technologies advertised at the conference are branded as being tested in police and military settings.
Some are fighting back against this AI surveillance push. Unions like the Retail, Wholesale and Department Store Union (RWDSU) have been incorporating clauses in new contracts that force employers to confer with workers before introducing new technologies to the shop floor, but this has yet to become the standard in an industry with such a low union rate.
At the state level, senators in Massachusetts, led by Sen. Dylan A. Fernandes, recently put forward a bill that would create new protections for workers by placing strict limits on monitoring, regulating automated decision systems, and allowing employees to refuse AI instructions.
“This technology has been adopted and applied at such a rapid pace that the law has not caught up,” Fernandes said. “We’re seeing a surge in technologies that can track keystrokes, monitor facial expressions, or automate discipline, often without transparency, oversight, or fairness.
“This bill is about drawing a clear line: Innovation cannot come at the expense of worker safety, or civil rights,” he added. “We don’t need to wait for a crisis or audit of abuses, this is about setting guardrails now to protect working people before harm becomes widespread.”
Alas, the federal government is ensconced with AI hype men, and as the Prospect has reported, Republicans’ One Big Beautiful Bill Act has a sneaky provision that would ban any state or local regulation of AI for ten years. Should it pass, retail employees and driverless car passengers alike better watch out.