A camera mounted on a grocery store shelf is watching you. It's analyzing your face, determining your age, your gender, possibly your race. It's cross-referencing that data with your purchase history, your browsing data, and information bought from data brokers about your income and shopping habits. An AI is calculating in real time the maximum price you personally will tolerate for the gallon of milk you just picked up. It updates the electronic price tag to display that number. You walk to the register, the price might change by the time you check out. The person next to you picking up the same milk sees a different price because the AI determined they would pay more. This is not a pitch for a Black Mirror episode. It's called the Kroger's Edge system enhanced display for grocery environment, which sounds less like a grocery store feature and more like a feature Microsoft would announce at a developer conference and then quietly discontinue 18 months later.

It's currently deployed in over 500 Kroger stores. Walmart is rolling out similar systems to 2,300 stores by 2026. Whole Foods and Amazon Fresh have them, too. Maryland just became the first state in the US to ban it. I'm L. I have a PhD in computer science and I analyze AI developments to understand what's actually happening beneath the hype. The technical term for this is dynamic pricing, but the more accurate term is surveillance pricing, which is when you know the marketing department lost the internal branding battle. Here's how the system works. Electronic shelf labels, small digital screens replacing paper price tags, let retailers change prices in less than seconds. The market for these devices was $1.85 billion in 2024 and is projected to reach $7.54 billion by 2033. That is a $7.5 billion industry built around making it easier to change the price of bread. Kroger introduced the system back in 2018. By 2024, the company partnered with a retail analytics firm called Intelligence Node, which is generally what they're called, Intelligence Node. Like they're sat around a conference table, and somebody said, "What if we named ourselves like an AI villain in a near-future dystopia?" And everyone just agreed that was the right move. Intelligence Node uses AI to provide dynamic pricing. Kroger also partnered up with Microsoft to develop the Edge Shelf system. Microsoft embedded cameras into the displays that use facial recognition to determine your age and gender. Kroger insists the cameras are only for personalized advertisements, not pricing, which is kind of reassuring. They're only watching your face to sell you things, not to charge you more. Totally different things, right? Kroger also partnered with a company called Cooler Screens, which replaces the transparent refrigerator doors with digital screens embedded with cameras. So, now you can see what's in the fridge. There's just a screen showing you ads for the yogurt you're trying to locate while a camera watches you squint at a display trying to figure out if they still have the Greek kind. The whole setup drew comparisons to Minority Report, a 2002 sci-fi film where Tom Cruise gets served ads because he's been implanted with another man's eyes. That's the cultural reference points we've arrived at. A Philip K. Dick story about surveillance and identity theft is now the grocery shop experience in 2026.

Back in August 2024, Senators Elizabeth Warren and Bob Casey sent Kroger CEO a letter with some questions. The key line was this, "Edge will allow Kroger to use customer data to build personalized profiles of each customer, and then use those profiles to determine how much price hiking each of us can tolerate, quickly updating and displaying the customer's maximum willingness to pay on the digital price tag. Your maximum willingness to pay, not the cost of production, not the market rate, your personal pain threshold for milk calculated by an algorithm in real time. The senators warned that, I'm quoting, "The price of basic household goods could surge based on the time of day, the weather, or other transitory events." The weather. Milk could cost more because it's raining, which makes sense if you think about it. People need milk when it rains, of course. Supply and demand, right? Except the supply hasn't changed, the demand hasn't changed, just the sky did a thing, so now the price is different. Totally makes sense. Representative Rashida Tlaib raised a different concerns entirely. Studies have shown that facial recognition technology is flawed and can lead to discrimination in predominantly black and brown communities. Facial recognition algorithms have higher error rates for darker skin tones. This is true. So, the system could misidentify black customers, build incorrect profiles, and charge them higher prices, not because of any malice, just because the AI is bad at faces nobody checked.

Kroger responded to all of this with a statement. To be clear, Kroger does not and has never engaged in surge pricing. Any test of electronic shelf tags is designed to lower prices for more customers where it matters most. To suggest otherwise is not true. Okay, so to recap, Kroger partnered with Intelligence Node for AI-powered dynamic pricing, Kroger partnered with Microsoft to embed facial recognition cameras through our stores, Kroger deployed electronic shelf labels that can change prices instantly across 500 stores, Kroger installed cameras in refrigerator doors, but the technology is exclusively for lowering prices. Yeah, I buy that. It's like installing a very elaborate knife sharpener in your kitchen and insisting it's only for cutting vegetables while standing in what appears to be a crime scene. In March 2026, Maryland Governor Wes Moore testified before the state legislature, I'm quoting, "If Miles and I were to go into a supermarket at the same time, we could pick up the exact same item and be charged a different price for it because they know that I'll pay it, which is a pretty clear articulation of the problem. Two people, same milk, different prices because an algorithm decided one of them is willing to tolerate more financial pain." Earlier this month, Maryland passed the Protection from Predatory Pricing Act banning dynamic pricing in grocery stores. The bill goes into effect in October this year. There's just one tiny issue. According to Justin Brookman, director of marketplace policy at Consumer Reports, the bill has no real enforcement. I'm quoting, "If you got caught doing it, all you need to do is stop once you get caught and then there can't be any fines. So, the punishment for using AI to algorithmically price gouge Maryland residents is we'd prefer if you didn't, please. No fines, no penalties, just a polite request to please stop." The Maryland Retailers Alliance, representing the stores being regulated, called it a workable framework, which makes perfect sense. Of course they think it's workable. They spent billions deploying surveillance infrastructure and the ban is please don't. That's extremely workable for them.

Meanwhile, income spent on food is at a 30-year high. Shoppers in Maryland are already experiencing price instability. Inflation is still above target. One told reporters, "You never know what the price is going to be. It fluctuates week on week." So, the retailers looked at existing price instability and basically thought, "Hmm, what if we made this happen faster? What if we made it happen in real time? What if we made it personalized?" Here's what's actually happening. Grocery stores build detailed profiles using loyalty cards, apps, browsing history, and data purchased from brokers. They categorize shoppers as interested in fitness and not price sensitive or expecting mother with a toddler. Then they sell that data to third parties. So, that's three revenue streams from one trip to buy milk. Surveillance, personalized pricing, and selling your data. The AI makes the scale. Kroger's system can show you one price at the shelf and charge a different price at checkout because the algorithm recalculated between those two moments. A human couldn't do this for thousands of customers simultaneously, but an AI definitely can. And because facial recognition has higher error rates for black faces, the system could systematically overcharge black customers due to misidentification. Not through intentional discrimination, just through algorithmic bias nobody bothered to actually check. Maryland's ban is the first, but it is structurally weak. No enforcement means no deterrent. And even if other states follow with stronger versions, this is ultimately a federal problem. It's a world problem. Kroger operates currently in 35 states. Walmart is basically everywhere. Banning dynamic pricing in Maryland while it's legal everywhere else just means Maryland cross-subsidized the system. The broader pattern is this ultimately. A 2021 UCLA study found that dynamic pricing can soften price competition, meaning retailers can all raise prices simultaneously without appearing to collude. The algorithm coordinates for them instead of just a couple of rich guys in a boardroom. So, what's actually going to happen next? Walmart is still deploying electronic shelf labels to 2,300 stores by 2026. Kroger has the system in 500 stores and expanding. The electronic shelf label market is projected to grow from $1.85 billion to $7.54 billion by 2033. The infrastructure is being deployed right now regardless of Maryland's ban. The cameras are being installed. The partnerships are being signed. Intelligence Node is still helping retailers calculate maximum willingness to pay. Microsoft is still embedding facial recognition in shelf displays. Maybe Kroger really is only using this to lower prices right now, but the capability exists. And when food costs are at historic highs and shareholders are demanding growth, maximum willingness to pay stops being a rhetorical concern and becomes operational reality. Maryland said no, the enforcement is weak, but the precedent matters a huge deal, especially right now with our current rising geopolitical tensions. One state decided that using AI to personalize grocery prices based on facial recognition and behavior profiling is not an acceptable business model. Whether that precedent spreads or gets buried depends on whether other states write bans with actual teeth. Because right now, the punishment for deploying surveillance capitalism in grocery stores is a polite request to please stop. And that is not a ban. That is more like a suggestion. And suggestions don't stop billion-dollar infrastructure rollouts, do they?