Your Price Was Set by AI: McDonald’s 21% Big Mac Gap and the Surveillance-Pricing Laws Starting 1 October (AI Trends, 29 September 2026)

admincybersecurity

What’s trending in AI on 29 September 2026: AI is quietly making decisions about ordinary people, starting with what they pay. A Reuters investigation found that McDonald’s runs a machine-learning engine that sets what it calls “the optimal price” at each of its nearly 14,000 US restaurants. In Fresno, California, one company-run store sold a Big Mac for $5.69 while another two miles away charged $6.89, a 21% premium. The timing is striking: on Thursday, 1 October, new surveillance-pricing laws take effect in Maryland and Connecticut. The same day brought two more stories about AI judging people: a UK police facial recognition trial that scanned 500,000+ faces for zero arrests, and a $4 billion housing AI company that keeps every binding decision with a human. This post explains what changed, where the legal line sits, and gives you a seven-step AI pricing audit to run before Thursday.

Key takeaways

  • AI now sets everyday prices. McDonald’s engine reportedly weighs millions of daily transactions, rivals’ menus and an estimate of how much each store’s customers are willing to pay.
  • The legal line is about data, not algorithms. Pricing from market conditions (dynamic pricing) is generally lawful. Pricing from who you are (surveillance pricing) is being banned or forced into disclosure.
  • The clock runs out Thursday. Maryland’s and Connecticut’s laws take effect 1 October 2026. New York already requires disclosure, New Jersey has signed a grocery ban, and 40+ bills are pending in 24+ states.
  • For your business: if any tool sets or recommends your prices, you need to know exactly what data feeds it, and whether competitors use the same tool.
Your price was set by AIA burger and a digital price tag wired to a neural network. Two store price cards read 5.69 dollars and 6.89 dollars, a 21 percent gap between two restaurants two miles apart.AI TRENDS · 29 SEPTEMBER 2026Your price wasset by AI.Same burger, two miles apart:$5.69 vs $6.89. New pricing laws start 1 Oct.Store A$5.69Store B · 2 miles away$6.89+21%$?.??delana.co
The price on the menu is increasingly the output of a model, not a manager’s decision.

Trend 1: McDonald’s lets AI set the price of a Big Mac

According to Reuters, McDonald’s pricing engine continuously analyzes millions of daily transactions across its nearly 14,000 US restaurants and recommends a price for each menu item at each location. Its inputs reportedly include menus scraped from nearby Wendy’s and Burger King stores, and, weighted heavily, an estimate of how much a particular store’s customers are willing to pay.

The clearest example Reuters found was in Fresno. Checking the McDonald’s app in September, it found one company-run store selling a Big Mac for $5.69 and another company-run store two miles away charging $6.89. Reuters is careful to say it could not confirm the engine caused that specific gap. But three franchisees said the tool has widened price differences between neighboring stores, and five said McDonald’s pressured them to use it.

Same sandwich, two miles apartBar chart: a Big Mac costs 5.69 dollars at one company-run McDonald’s in Fresno and 6.89 dollars at another two miles away, a 21 percent premium. The pricing engine reportedly weighs millions of daily transactions, nearby competitor menus and each store’s estimated willingness to pay.Same sandwich, two miles apartBig Mac price in the McDonald’s app at two company-run stores in Fresno, CA (Sept 2026)$0$1$2$3$4$5$6$7Store A$5.69Store B2 miles away$6.89+21%What the pricing engine reportedly weighsMillions of dailytransactions, ~14,000 storesScraped menus of nearbycompetitorsEstimated willingness to payof each store’s customersSource: Reuters investigation, 29 Sept 2026 (via AI Weekly). Reuters could not confirm the engine caused this specific gap. · delana.co
A 21% gap for the same sandwich. The engine’s most sensitive input is an estimate of what each store’s customers will tolerate.

“A tool, not a mandate”?

McDonald’s describes its pricing portal as a tool that gives franchisees restaurant-specific recommendations, not a requirement. The reporting complicates that. Since January, McDonald’s business standards have required franchisees to engage constructively with its approved pricing consultant and tools, and on the August earnings call CEO Chris Kempczinski linked “pricing non-compliance” to conversations with owners who hadn’t adopted the chain’s value-menu guidance.

Two details should interest anyone who runs a business. First, the portal itself reportedly warns franchisees that they “may be competitors” and face antitrust risk, an admission that a shared pricing brain across independent owners is legally sensitive. Second, a lawsuit by Connecticut franchisee George Michell alleges the tool suggested roughly $18 for a Big Mac meal at a turnpike location. McDonald’s disputes the suit, and Reuters could not verify that recommendation.

Where the legal line actually sits

Regulators are not banning algorithms. They are separating two practices that can run on the same software:

  • Dynamic pricing reacts to the market: demand, time of day, season, inventory, local costs. Everyone at the same place and time sees the same price. Airlines and hotels have done this for decades, and it is generally lawful.
  • Surveillance pricing (also called personalized algorithmic pricing) reacts to the person: browsing history, location, device, purchase history, inferred income or family size. Two shoppers can see two prices for the same item at the same moment. This is the target of nearly every new law and enforcement action.
Dynamic pricing vs surveillance pricingComparison. Dynamic pricing sets prices from market signals such as demand, time, season and local costs, so everyone in the same place sees one price. Surveillance pricing uses personal data such as browsing, location, device and inferred income, so two shoppers can see two prices; it is banned or requires disclosure in a growing list of states. A third risk is competitors sharing one pricing algorithm, as in the DOJ RealPage settlement.The line regulators are drawingSame algorithm, very different legal risk. The question is what data sets the price.Dynamic pricingPrice follows the market• Demand, time of day, season, stock• Store costs and local competition• Everyone in the same place sees one priceExamplesAirline fares, surge delivery fees,end-of-day markdownsSurveillance pricingPrice follows who you are• Browsing, location, device, purchase history• Inferred income, family size, urgency• Two shoppers, same moment, two pricesStatusBanned or disclosure-only in a growinglist of states; FTC and state AGs activeThird risk: one algorithm shared by competitors.The DOJ’s RealPage settlement (Nov 2025) limited data sharing among clients of a common pricing tool.Sources: Holland & Knight (Aug 2026), DOJ · delana.co
What decides your legal exposure isn’t whether you use AI to price. It’s what data the AI is allowed to see.

Based on what Reuters describes, McDonald’s engine looks more like store-level dynamic pricing than personal surveillance pricing: it estimates what a location’s customers will bear, not what you will bear. That distinction matters, but it isn’t a clean bill of health. A third risk sits between the two columns: many competitors feeding one shared pricing algorithm. That was the heart of the Justice Department’s case against rental-pricing software maker RealPage, settled in November 2025 with limits on how client data can be pooled. It is also why McDonald’s own portal reportedly warns franchisees about antitrust exposure.

The surveillance-pricing laws: what takes effect and when

Pricing by algorithm has gone from a niche privacy topic to a bipartisan consumer issue in under two years. Here is where the rules stand, according to law-firm trackers from Holland & Knight, Wilson Sonsini and others:

  • New York (in force since November 2025): the Algorithmic Pricing Disclosure Act requires businesses to tell consumers when a price was set by an algorithm using their personal data. It survived a First Amendment challenge in federal court.
  • Maryland (1 October 2026): the Protection From Predatory Pricing Act bans surveillance pricing by larger food retailers and delivery platforms.
  • Connecticut (1 October 2026): bans surveillance pricing and requires a consumer-facing label: “THIS PRICE WAS INCREASED BY A PRICE SETTING DEVICE USING YOUR PERSONAL DATA.”
  • New Jersey (signed 23 July 2026): the Fair Price Protection Act bans surveillance pricing for groceries, in stores and online, with penalties up to $50,000 per violation, treble damages and a private right to sue. It takes effect about a year after signing.
  • New York, round two: the One Fair Price Act, which would replace disclosure with an outright ban, passed the Legislature in June. As of the latest law-firm updates, it awaited Governor Kathy Hochul’s decision.
  • Federal: the FTC’s 14 April 2026 advance notice of rulemaking asks whether delivery platforms disclose personalized prices, and a Senate Judiciary subcommittee held a hearing on AI surveillance pricing in August.
The surveillance-pricing clockTimeline: November 2025, New York’s Algorithmic Pricing Disclosure Act in force. 14 April 2026, FTC advance notice covers personalized pricing disclosure. June 2026, New York’s One Fair Price Act passes the Legislature. 23 July 2026, New Jersey signs a grocery surveillance-pricing ban with penalties up to 50,000 dollars per violation. 1 October 2026, Maryland and Connecticut laws take effect. More than 40 bills are pending in over 24 states.The surveillance-pricing clockKey US rules on AI and algorithmic pricing, 2025–2026Nov 2025New YorkAlgorithmic PricingDisclosure Act in force14 Apr 2026FTCAdvance notice coverspersonalized pricingJun 2026New YorkOne Fair Price Actpasses Legislature23 Jul 2026New JerseyGrocery ban signed;$50K per violation1 Oct 2026MD + CTSurveillance-pricinglaws take effectTHIS THURSDAY40+ more bills pending in 24+ states · NJ law takes effect about a year after signingSources: Holland & Knight, Wilson Sonsini, Consumer Finance Monitor · delana.co
States are moving faster than Washington. Two laws take effect on 1 October, and dozens more are in the pipeline.

Most enacted laws focus on groceries and delivery for now, so a burger chain is not the obvious first target. But state attorneys general don’t need a pricing-specific statute to act; they can use existing unfair-and-deceptive-practices laws, and several already have. New York’s attorney general pressed a delivery platform over pricing experiments in January, and California’s launched a surveillance-pricing sweep the same month.

Trend 2: two very different ways to let AI decide about people

Pricing wasn’t the only AI-judges-people story on 29 September. Two others make a useful pair, because they show opposite design choices.

Scan everyone, measure later. Freedom of information documents reported by The Guardian show that British Transport Police’s six-month live facial recognition trial at London railway stations scanned more than 500,000 faces between February and July 2026. It produced one alert, which was a false positive, and no direct arrests, while costing £320,786 and nearly 100 hours of officer time across 18 deployments. The pilot has since expanded to selected London Underground stations and runs until November.

Automate the work, keep the verdict human. EliseAI, which automates leasing and resident services for landlords (and patient calls for physician groups), raised $350 million at a $4 billion valuation, Fortune reported. It says its platform now powers roughly one in six US apartments and handles about 5 million calls a month. Its CEO, Minna Song, told Fortune that anything ending in a binding decision, such as approving or denying an application, sending a formal notice or agreeing lease terms, still goes through a person, and that the AI says so when it lacks information rather than guessing.

Two ways to let AI decide about peopleStat tiles. British Transport Police’s live facial recognition trial scanned more than 500,000 faces, produced one alert that was wrong, led to zero direct arrests and cost 320,786 pounds and about 100 officer hours. EliseAI’s AI serves about 1 in 6 US apartments and handles about 5 million calls a month, but approvals, denials, formal notices and lease terms still require human sign-off.Two ways to let AI decide about peopleBoth stories broke on 29 September 2026Scan everyone, check the value laterBritish Transport Police live facial recognition trial, London stations, Feb–Jul 2026500,000+faces scanned1alert, and it was wrong0direct arrests£320,786cost, ~100 officer hoursAutomate the work, keep the verdict humanEliseAI, leasing and resident-services AI, raised $350M at a $4B valuation1 in 6of US apartments~5Mcalls handled a monthHuman sign-offon approvals, denials, formal notices, lease termsSources: The Guardian (FOI documents), Fortune · delana.co
Same question, opposite answers: does the AI decide, or does it prepare the decision for a person?

The lesson carries straight back to pricing. An AI that recommends a price for a manager to approve, with its inputs logged, is a very different legal and ethical object from one that silently sets prices per customer. And a system that scans half a million people to produce one wrong alert is a reminder to measure what an AI deployment actually delivers before scaling it. We made the same point about camera-equipped devices in our AI glasses workplace policy.

Also trending in AI today

  • OpenAI DevDay 2026 opens today in San Francisco at 10 a.m. PT, with more than 20 announcements teased and the company under pressure over recent agent safety incidents. We previewed it in the AI week ahead.
  • OpenAI’s revenue run rate is nearing $70 billion, up more than 70% since the start of the third quarter, Axios reported.
  • Anthropic’s IPO filing shows at least $518 billion of planned AI infrastructure spending over a decade, about 80% of it non-cancelable, per Reuters. For the chipmaker side of the money story, see this morning’s AMD and World Labs breakdown.
  • Claude had a partial outage of about 40 minutes (14:21 to 14:59 UTC) affecting claude.ai, the API and Claude Code, according to Anthropic’s status page. A good reminder to have a fallback for any AI tool your team depends on.
  • Google appealed an EU Digital Markets Act order that would force Android to open up to rival AI assistants, Bloomberg reported.

What this means for your business

  • You may already be using AI pricing without calling it that. E-commerce platforms, booking engines, delivery apps, ad-driven discount tools and loyalty programs increasingly adjust prices automatically. The law cares what data they use, not what the feature is called.
  • Your vendor’s data is your liability. Outsourcing pricing to a tool does not outsource responsibility. If the tool pools data across competitors, you may share their antitrust exposure.
  • Location is a gray zone. Store-level or regional prices are normal. Using a person’s real-time location or inferred income to raise their price is exactly what the new laws target.
  • Documentation is your defense. Lawyers tracking these laws say companies that can show they train staff on the dynamic-versus-surveillance distinction, and can justify every price difference, are in a much stronger position when a regulator calls.

The 7-step AI pricing audit (run it before 1 October)

This takes a few hours for most small and mid-sized businesses. It isn’t legal advice, but it tells you whether you need some.

  1. Inventory every price-setting tool. List every system that sets, adjusts or recommends a price or discount: your store platform, booking system, delivery partners, promotions engine and any “smart pricing” add-on.
  2. Map the inputs. For each tool, write down exactly what data it uses. Flag anything personal: account history, browsing, device type, precise location, inferred demographics.
  3. Classify each tool. Market signals only means dynamic pricing. Any personal data that changes an individual’s price means potential surveillance pricing. Treat “we’re not sure” as the second category until proven otherwise.
  4. Check your map against the states you sell into. Start with New York, Maryland, Connecticut and New Jersey, then check where you have customers among the 24+ states with pending bills.
  5. Ask your vendors three questions. What data trains and drives the model? Is data pooled across clients who compete with us? Will you indemnify us if the tool breaks a pricing law?
  6. Keep a human in the loop for outliers. Set a threshold (for example, any price more than 15% above your baseline) that requires a person to approve, and log the reason.
  7. Prepare the disclosure and the paper trail. Know where a Connecticut- or New York-style label would appear if you need one, and keep dated records of why each price difference exists.

Frequently asked questions

What is surveillance pricing?

Surveillance pricing, also called personalized algorithmic pricing, is when a business uses a customer’s personal data, such as browsing history, location, device, purchase history or inferred income, to set an individual price. It means two people can be charged different prices for the same product at the same time.

Is dynamic pricing legal?

Generally, yes. Dynamic pricing changes prices based on market conditions like demand, time of day, season or inventory, and everyone in the same situation sees the same price. It is not the main target of the new state laws, although shared pricing algorithms among competitors can raise antitrust issues.

Does McDonald’s use surveillance pricing?

Reuters describes a store-level system that recommends prices for each restaurant using transaction data, competitor menus and an estimate of each store’s customers’ willingness to pay. That is closer to location-based dynamic pricing than individual surveillance pricing. Reuters did not report evidence of individualized prices, and could not confirm the engine caused the specific Fresno price gap.

Which surveillance-pricing laws take effect on 1 October 2026?

Maryland’s Protection From Predatory Pricing Act and Connecticut’s surveillance-pricing ban, which also requires a disclosure label when personal data raised a price. New York’s disclosure law has applied since November 2025, and New Jersey’s grocery ban was signed in July 2026 and takes effect about a year later.

Do these laws apply to small businesses?

Scope varies by state. Maryland’s ban focuses on larger food retailers and delivery platforms, while others apply more broadly, and state attorneys general can also use general consumer-protection law. If any tool you use sets prices from customer data, check the specific rules in each state where you sell, and talk to a lawyer if you’re unsure.

Sources