You Are Not the User

On September 25 the chief executive of Walmart published a letter to his customers. “We price the product, not the person,” John Furner wrote. He promised that a shopper’s income, shopping history, or urgency would not change a price.

Ten days later Walmart published a second letter. It was addressed to Lindsay Owens, who runs a progressive research group in Washington called Groundwork Collaborative and had just published a book on pricing technology. The letter asked her to correct the record. The same day her group released a report on Walmart’s patents.

One of those patents was granted in 2023. It covers changing the price on a store’s electronic shelf label according to what a shopper has already put in the cart. The text gives an example: “if a customer has tuna fish, they may be offered a different price for mayonnaise.” Walmart replied that a patent describes what could be built, and that “a patent isn’t a plan.” It said tuna in a cart does not change what it charges for mayonnaise, and that it will not renew the patent.

What the record supports

This essay checked what it could. Walmart has said in writing, twice, that it does not set prices by person and will not. This essay found no one who has shown otherwise, and it accepts that record. The fact-checking site Snopes examined one widely shared claim in September and reported no credible evidence that Walmart uses customer data to set individual prices.

Several things are documented. Walmart is replacing paper price tags with digital labels. It runs an AI shopping assistant called Sparky, and its executives have told investors that customers who use it spend about 40 percent more per order. The company says that figure does not mean anyone paid more for the same item. Its privacy notice says it uses license plate readers on its properties for security. It holds the patents, and two have the words “dynamic pricing” in the title.

Some claims did not hold up when checked. This essay found no evidence that parking lot cameras are used to set prices, or that a price-setting shopping cart is being tested.

Room for doubt

Accepting the record does not close the question.

Walmart is right that a patent is not a plan. Companies file many patents and use few. But a patent is also a choice. It costs money, and it means someone inside the company thought the idea was worth owning.

These ideas were more than a single filing. One of the dynamic pricing patents grew out of an application made in 2014. It was granted in 2021, and it covers adjusting an online price using, among other things, a shopper’s past behavior. In July 2024 Walmart paid the fee to keep it in force. The shelf label application of 2018 described pricing by what is in a shopper’s cart, but it did not ask for protection of that idea. By the time the patent was granted in 2023, the cart was at the center of what it protects. Walmart said it would not renew that patent the day after the Groundwork report appeared. This essay found no such statement about the others.

There is also the question of incentive. A pledge asks a company to leave a profitable tool unused, while its executives tell investors that the shopping assistant raises what customers spend. Phil Lempert, a longtime grocery analyst, wrote in October that he believes Furner. He added that the promise will be worth only as much as the next chief executive decides.

And there is the question of what can be seen. On a phone or a website, each shopper sees only their own screen. If a seller did price the person there, the buyer could not tell. So the fact that no one has caught it says less than it seems to.

Opinion 1. This essay’s opinion is that skepticism about intentions is fair here. Skepticism is not an accusation. It is the ordinary response to a party that has the tools and the incentive, and whose conduct is hard to watch from outside.

Not only Walmart

The same doubt reaches other large institutions. The evidence comes in three strengths, and it helps to keep them apart.

Some conduct has been shown. In December 2025 Consumer Reports, working with Groundwork and a third group, published the results of five tests at stores in four cities. In each test, a group of volunteers added the same items to their Instacart carts, from the same store, at the same time. More than 400 volunteers took part. About three quarters of the items appeared at more than one price, and the gap on a single item reached 23 percent. Instacart said the tests were random, and it ended them two weeks later. Governments belong on this list. The Dutch tax authority used algorithms to pick families for fraud checks, and it counted a second nationality against them. Thousands of families were wrongly accused, and the government resigned over it in 2021.

Some conduct has been charged or announced. In 2024 the Justice Department sued a software company called RealPage. It alleged that landlords gave the company their private rent data and that its algorithm sent back recommended prices. “Training a machine to break the law is still breaking the law,” the deputy attorney general said. RealPage agreed to a settlement in 2025 without admitting liability. In July 2025 Delta told investors it planned to use AI pricing across a fifth of its domestic network by the end of that year. Three senators wrote to Delta with concerns that fares would be set by person. Delta answered that it had never used, tested, or planned such fares.

And some is left open. That is where Walmart belongs. The tools exist, the pledge exists, and an outsider cannot readily check which one the company follows.

The amplifier

None of this conduct began with AI. For most of history a price was whatever a buyer and a seller could agree on. The Quakers were among the first merchants to commit to one fixed price for everyone, and they did it on moral grounds. Sellers went on trying to learn what customers would buy and what they would pay. In 2004 The New York Times reported that Walmart’s data showed sales of strawberry Pop-Tarts rising to seven times the normal rate ahead of a hurricane. In 2012 The Wall Street Journal found the Staples website showing different prices for the same stapler, depending on where it estimated the shopper was.

What AI changes is the shape. A clerk could judge one customer at a time. A pricing system can do it for every shopper and every item. In January 2025 the staff of the Federal Trade Commission reported that firms selling pricing tools had at least 250 clients. It said details as small as a shopper’s mouse movements could be used to tailor a price.

This first effect has a name. In 2015 the computer scientist Kentaro Toyama called it the law of amplification: technology’s main effect is to amplify human forces that are already there. That includes the good ones. The kind of engine that could raise a price can also lower one.

The muffler

The second effect gets less attention. The same systems can make it harder to see who is acting.

One kind hides who is influencing a purchase. Instacart’s own website said shoppers “are not aware that they’re in an experiment.” It took groups of volunteers shopping at the same moment to find it, because one shopper sees one price. In the old haggling, both sides knew a negotiation was going on. Here only one side knows.

The other kind hides who did the work. In January 2025 the Securities and Exchange Commission found that a company called Presto had claimed its AI took drive-through orders without human order-takers. In fact people stepped in on the vast majority of orders. The reverse also happens. In October 2025 Deloitte agreed to refund part of its fee to the Australian government for a report that cited research papers that do not exist. The corrected version disclosed that an AI tool had been used.

Others have named parts of this. Dan Davies calls a system that absorbs blame an “accountability sink.” Frank Pasquale wrote of a “black box society” in 2015. Neither effect is this essay’s discovery. Its contribution is to put the two side by side and to ask what each one calls for.

The word muffler is chosen with some care. A muffler does not weaken an engine. The engine runs as hard as before, and the people nearby hear less of it. The muffler is also why proof comes late. A practice that is hard to see is hard to prove.

What the three answers leave out

“Wrong Date, Wrong Institution, Wrong Scale” answered three people who spoke about AI on Bloomberg: Michael Sandel, Daniel Diermeier, and Meghan Sullivan. This essay keeps its respect for all three and adds one observation. Much of the segment concerned people who use AI, or who decide whether to.

A shopper at the shelf is not the user. The store is. The same is true of a renter, a passenger, and a family applying for a benefit.

Opinion 2. That changes how far each answer reaches. Sandel fears a change in what it means to be human, and the earlier essay agreed that something human is being lost. This essay’s opinion is that the harm described here is nearer. It has dates and named parties, and it can reach people who never open a chatbot. Diermeier wants students who can think for themselves. A shopper can think very well and still not see how a price was set. Sullivan came closest. She said ordinary people are not helpless. But how a store sets its prices is not the shopper’s choice to make. It belongs to the store, and to whoever writes the rules for stores.

Where the rules are being written

Some of those rules now exist. Maryland’s ban on using a shopper’s personal data to charge more for groceries took effect on October 1. Connecticut and New Jersey have enacted bans of their own. New York’s legislature passed one in June, and the governor had not signed it when this was written. In Washington, a Republican, Josh Hawley, chaired a Senate hearing on the practice in August. A Democrat on the panel, Richard Blumenthal, described “a bipartisan feeling of urgency.” A bill in the House is led by a Democrat and a Republican.

This essay should not oversell any of it. No federal bill has passed. The three state laws come from states led by Democrats, and Colorado’s Democratic governor vetoed a similar bill. Consumer Reports said Maryland’s law has too many loopholes to protect shoppers. Grocers’ trade groups say their members do not price by person, and they warn that broad rules could reduce loyalty discounts.

What matters here is the target. These laws do not regulate a model, a university, or a member of the public. They regulate what an institution may do to a person with a machine, and they give someone outside the institution the power to check. That is the scale the earlier essay asked for.

Builders and users

Two arguments about AI rules are often treated as one. They should be kept apart.

The first is about the labs that build the largest models, and how tightly to regulate them. That question is not easy. Anyone who says it is easy is not counting what it would cost, at home and abroad, to slow AI too much. The earlier essay noted two facts that bear on this. Reports cited in the Bloomberg segment put AI investment at about 2 percent of the American economy, and China has set national targets for AI adoption. Too little restraint has costs of its own. This author welcomes that argument and thinks it has to happen. He does not yet have an opinion on where the line belongs. Wherever it is drawn, it will have costs and benefits. No line is free.

Opinion 3. Whatever is decided there, this essay’s main point is about where the burden of conduct should sit. It should not sit mostly on individuals. That asks people to regulate, one at a time, institutional conduct they cannot see. It should not sit mostly on the labs either. The labs are coming to look like utilities. Sam Altman, who runs OpenAI, has said he sees a future where intelligence is “a utility, like electricity or water.” A power company answers for the safety of its own lines. It does not answer for what every customer does with the current. To put too much of the burden on the labs is to ask them to hold back development, in ways that could work against national security and against legitimate uses of the tools.

Opinion 4. That leaves the second argument, which is about the governments and companies that use AI on other people. It does not matter whether they built the model or bought it from a lab. Here this essay does have an opinion. The burden belongs with them, and the time for legislation is now, before the harm is proved in every case.

Opinion 5. The reason is simple. “The Durable Organization,” another essay by this author, argues that change put off does not go away. It arrives later as one large change, under pressure. This essay’s opinion is that rules work the same way. The longer institutions are allowed to amplify their reach and muffle their conduct in unethical or manipulative ways, the harder that will be to unwind. Instacart ended its price tests two weeks after they became public. By its own account, those tests involved a small number of retailers. A practice that has spread across an industry, with revenue depending on it, will be much harder to end.

These rules are not free either. Drafted badly, they could reduce ordinary discounts, as the grocers warn. But they bar particular conduct, and a company that already keeps its pledge gives up little when the pledge becomes law.

Two questions

The amplifier and the muffler give anyone two questions to ask about a use of AI. What is it making stronger? What is it making harder to see?

What is it making stronger? What is it making harder to see?

The answers call for different remedies. Where a practice has grown too strong, the remedy is a limit, such as a ban on setting a price from personal data. Where something has gone out of sight, the remedy is to bring it back into view. That can mean a label on the price and a record an outsider can inspect. It also means a named person who answers for the result.

“The Truth Will Set You Free,” another earlier essay, offered a test for any use of these machines. Does it leave people able to understand, to check, and to answer for a decision? A muffled system fails the second part, and often the third. That essay also argued that announcements rarely change what people are rewarded for doing. A pledge is an announcement. Walmart’s letter is an institution binding itself in public, and that deserves credit. A promise that outsiders cannot readily check is not yet a rule.

What this essay is against

This essay is not against capitalism. It is not against profit, low prices, or the tools. The same engines can mark down a price or catch a fraud.

Opinion 6. It is against one thing. That is any institution, a company or a government, using AI to amplify unethical conduct while muffling it for the people it is used on. One fair test of conduct is whether it could be announced to the people it affects. Instacart’s price tests ended two weeks after they became public.

That conduct is old. AI changes its shape. It reaches more people, and it is harder to see. Proof exists in some of the cases above and not in others. Where there is none, this essay claims only that the doubt is legitimate, and that the answer to legitimate doubt is a way to check.

A scapegoat

Opinion 7. In an interview published in early October, Altman told Vanity Fair that people anxious about AI “need a sink for the anxiety, a scapegoat.” He is probably right that anyone in his job would be treated the same way. This essay’s guess is that the muffler is part of the reason. When people cannot see who is acting on them, the blame tends to go to a person they can name, or to the machine.

An institution can point at the machine too. In 2024 Air Canada argued before a Canadian tribunal that it was not liable for what the chatbot on its own website told a customer. The tribunal called that a “remarkable submission” and held the airline responsible.

So the better aim is neither Altman nor “AI.” It is the company or the agency that uses a system on people, together with the rule that says what it may do and who answers for it. Whoever uses these tools on other people should have to do it in the open.

References

1. On Walmart’s two letters and the Groundwork report: John Furner, “A Letter from Our CEO,” Walmart, 25 September 2026. Dan Bartlett, “Correcting the Record: We Price Products, Not People,” Walmart, 5 October 2026. Lindsay Owens and Elizabeth Pancotti, “The Walmart Watchtower: Patents Speak Louder Than Pledges,” Groundwork Collaborative, 5 October 2026. Dani James, “Walmart defends pricing practices, again,” Retail Dive, 6 October 2026.

2. On what is documented about Walmart: Snopes, “Did Walmart update its privacy policy to allow surveillance pricing?” 10 September 2026. Walmart, “Digital Shelf Labels: A Modern Tool for Every Day Low Pricing,” 4 September 2026. Walmart, corrected transcript of its session at the Goldman Sachs Global Retailing Conference, remarks by Dave Guggina, 15 September 2026. Walmart, “Walmart Automated License Plate Readers (ALPR) Privacy Notice,” updated 16 February 2026.

3. On Walmart’s patents: Walmart Apollo, LLC, U.S. Patent 11,687,872 B2, “System and method for remote controlling of electronic shelf labels,” filed 19 March 2018, granted 27 June 2023. Walmart Apollo, LLC, U.S. Patent Application Publication 2018/0276603 A1, “System and method for remote controlling of electronic shelf labels,” published 27 September 2018. Walmart Apollo, LLC, U.S. Patent 10,896,433 B2, “Dynamic pricing systems and methods,” granted 19 January 2021. The patent family’s earliest priority date is 30 April 2014. The legal events list a fourth-year maintenance fee paid on 17 July 2024. Walmart Apollo, LLC, U.S. Patent 11,429,992 B2, “Systems and methods for dynamic pricing,” filed 27 November 2018, granted 30 August 2022. Walmart Apollo, LLC (originally Wal-Mart Stores, Inc.), U.S. Patent Application Publication 2018/0060884 A1, “Vehicle Suspension Measurement System and Associated Methods,” published 1 March 2018. The legal events list a final rejection on 11 May 2020 and abandonment on 5 December 2020.

4. On the claims that did not hold up, and on the pledge: Chris Dickerson, “Retail expert: Claims in ‘fair price’ book unfounded, impractical,” Legal Newsline, 2 October 2026. Phil Lempert, “Who’s Right About Walmart’s Pricing?” The Robin Report, 7 October 2026.

5. On the Instacart tests: Katie J. Wells, Lindsay Owens, Angel Han, and Alan Smith, “Same Cart, Different Price: Instacart’s Price Experiments Cost Families at Checkout,” Groundwork Collaborative and Consumer Reports, 9 December 2025. Consumer Reports, “New Report Exposes Instacart’s Hidden Price Games,” press release, 9 December 2025. NBC News, “Instacart to halt ‘price tests’ amid scrutiny of its AI tool for retailers,” December 2025.

6. On the Dutch childcare benefits case: DutchNews.nl, “Tax office fined €2.7m for discriminating against dual national families,” 8 December 2021. European Parliament, “The Dutch childcare benefit scandal, institutional racism and algorithms,” question for oral answer O-000028/2022, 28 June 2022.

7. On RealPage and Delta: U.S. Department of Justice, “Justice Department Sues RealPage for Algorithmic Pricing Scheme that Harms Millions of American Renters,” press release, 23 August 2024. Jeffrey C. Bank and others, “DOJ Settles Its Algorithmic Price-Fixing Case Against RealPage,” Wilson Sonsini Goodrich & Rosati, 1 December 2025. David Shepardson, “Delta plans to use AI in ticket pricing draws fire from US lawmakers,” Reuters, 22 July 2025, as carried by Yahoo Finance. Marlon Hyde, “Delta Air Lines plans to use more AI to price tickets, but what data will it use?” WABE, 31 July 2025.

8. On prices before AI, and on today’s pricing tools: Bronson Arcuri and Benjamin Naddaff-Hafrey, “The Price Tag Hasn’t Always Existed, It Had To Be Invented,” NPR Planet Money, 28 February 2018, as carried by WUWM. Constance L. Hays, “What Wal-Mart Knows About Customers’ Habits,” The New York Times, 14 November 2004. The Wall Street Journal, “Websites Vary Prices, Deals Based on Users’ Information,” December 2012. Federal Trade Commission, “FTC Surveillance Pricing Study Indicates Wide Range of Personal Data Used to Set Individualized Consumer Prices,” press release, 17 January 2025.

9. On amplification, blame, and the black box: Kentaro Toyama, Geek Heresy: Rescuing Social Change from the Cult of Technology, PublicAffairs, 2015. Dan Davies, The Unaccountability Machine: Why Big Systems Make Terrible Decisions and How the World Lost Its Mind, Profile Books, 2024. Frank Pasquale, The Black Box Society: The Secret Algorithms That Control Money and Information, Harvard University Press, 2015.

10. On who did the work: U.S. Securities and Exchange Commission, “SEC Charges Restaurant-Technology Company Presto Automation for Misleading Statements About AI Product,” 14 January 2025. Associated Press, “Deloitte to partially refund Australian government for report with apparent AI-generated errors,” 7 October 2025, as carried by the Jamaica Gleaner.

11. On the Bloomberg segment and earlier essays by this author: Bloomberg, report by Chrystia Freeland with Michael Sandel, Daniel Diermeier, and Meghan Sullivan, as posted by RealClearPolitics under the headline “Philosopher Michael Sandel: The Real AI Threat Is Changing What It Means To Be Human,” 5 October 2026. John Elijah, “Wrong Date, Wrong Institution, Wrong Scale,” October 2026. John Elijah, “The Truth Will Set You Free,” October 2026. John Elijah, “The Durable Organization,” October 2026.

12. On the state laws: Consumer Reports, “Consumer Reports statement on the signing of Maryland’s Protection from Predatory Pricing Act into law,” 28 April 2026. The Wise Marketer, “Surveillance Pricing, State by State: The Complete Legislative Tracker,” 11 September 2026. Electronic Privacy Information Center, “Connecticut Is Second State to Enact Surveillance Pricing Ban,” 4 June 2026. The Shelby Report, “NJ Gov. Signs Surveillance Pricing Ban Despite Industry Objections,” 24 July 2026. Spectrum News NY1, “NYC businesses urge Hochul to back changes to bill banning ‘surveillance pricing,’” 12 August 2026. New York State Senate, Senate Bill S8623B, 2025-2026 session, “Prohibits the use of surveillance pricing,” bill record as read on 8 October 2026. ICSC, “Surveillance Pricing Legislation Gains Momentum as States Wrap Up 2026 Sessions,” 9 July 2026.

13. On Congress and the grocers’ trade groups: Consumer Finance Monitor, “Senate Judiciary Hearing Reveals Bipartisan Support for Federal Action on AI-Driven ‘Surveillance Pricing,’” 14 August 2026. Rev, “Senate Hearing on Surveillance Pricing,” transcript of the 4 August 2026 hearing of the Senate Judiciary Subcommittee on Crime and Counterterrorism. H.R. 8895, No Rigged Grocery Prices Act, 119th Congress, introduced 19 May 2026 by Rep. Josh Gottheimer (D-N.J.) with Rep. Michael Lawler (R-N.Y.). Lorrie Griffith, “FMI, NGA Push Back After Senate Hearing On AI Surveillance Pricing,” The Shelby Report, 6 August 2026. Joshua Mitchell, “NJ Grocery Shoppers Can Sue Retailers for Surveillance Pricing Under New Law,” Tech Times, 25 July 2026.

14. On China’s targets and on intelligence as a utility: State Council of China, “China issues guideline to accelerate ‘AI Plus’ integration across key sectors,” 27 August 2025. AJ Dellinger, “Sam Altman Says Intelligence Will Be a Utility, and He’s Just the Man to Collect the Bills,” Gizmodo, 12 March 2026.

15. On the scapegoat remark and the Air Canada ruling: RealClearPolitics, “Sam Altman on AI Fear: ‘People Need a Scapegoat,’ ‘I Take It Somewhat Personally,’” video and transcript of an interview with Mark Guiducci of Vanity Fair, 7 October 2026. Moffatt v. Air Canada, 2024 BCCRT 149, British Columbia Civil Resolution Tribunal, 14 February 2024. Proctor (Queensland Law Society), “It wasn’t us, the chatbot did it,” February 2024.