AI Regulation Tracker / Consumer protection
New Jersey drew the line at surveillance pricing on necessities
The Fair Price Protection Act, A4523, was signed on July 23, 2026. It bars businesses from using personal data to charge one shopper more than another for the same necessity, and it hands enforcement to the Attorney General.
What exactly did New Jersey sign?
The Governor's office put the operative sentence in the release itself. It is the cleanest available statement of what the statute reaches.
The Act prevents businesses from using personal information ... to charge different prices for identical products based on what an algorithm predicts a shopper is willing or able to pay.Office of the Governor, State of New Jersey, signing release, July 23, 2026
The release lists the personal information it has in mind: online activity, location, purchasing history, and other collected data. It also gives the state's working definition, describing surveillance pricing as the use of consumer data to set prices for individual shoppers, often through large-scale data collection and artificial intelligence.
Read those passages together and the structure of the prohibition falls out. Three elements have to line up. Personal information about a particular shopper. An algorithmic prediction of what that shopper will tolerate. A resulting price difference on an identical product. Remove any one and you are outside the description the state itself published.
What is surveillance pricing, and what is ordinary dynamic pricing?
This is the whole compliance question, and most people have it blurred.
Dynamic pricing changes the price of a good for everyone at once. Airfare rises as the cabin fills. A ride costs more in the rain. Produce is marked down at nine in the evening. The variable is time, inventory, cost or aggregate demand, and two customers in the same store at the same moment see the same number.
Surveillance pricing, as New Jersey describes it, changes the price for a person. The input is data about that individual and the output is a number calculated for that individual, on an identical product, at the same moment. Two customers see different numbers because the system reached different conclusions about their wallets.
The popular use of the term is looser. Commentary sweeps in any opaque pricing algorithm, including pure demand models with no personal data in the pipeline. The state's language is narrower. It is anchored on personal information and on identical products, and it carves out loyalty programs and ordinary discounts, which are also price differences between individuals but ones the shopper opts into and can see. A discount you claim with a card is a price difference. A discount an algorithm decides you will not notice is one too. New Jersey treated the first as normal commerce and the second as the target.
Which goods does the ban actually cover?
The measure focuses on groceries and other necessities, not all goods sold in New Jersey. This is not a general prohibition on personalised pricing across the New Jersey economy, and reading it that way sends a retail team into the wrong scoping exercise.
The Act focuses on necessities. Groceries sit squarely inside that category, and the Governor's framing puts food at the centre of the concern. The Act defines which goods count as necessities. This entry does not enumerate that definition because the signing release does not provide it and the bill text was not read for this purpose. Consult the enrolled text of A4523 for the complete statutory definition.
For a retailer that produces a two-part scoping question. Does the personalisation engine set individual prices from individual data? And does the catalogue include goods the Act defines as necessities? A grocery chain answers yes to the second almost by definition. A general merchandise marketplace has to run the definition against its own categories.
Why does this entry not state an effective date?
Because the primary source does not, and a compliance date is not something to guess at.
The Governor's release announces the signing on July 23, 2026 but does not state when the prohibition becomes operative. New Jersey statutes commonly carry an operative-date clause at the end of the bill, and the pattern varies: immediate on enactment, the first day of a later month, or a fixed date months out. The sibling FAIR Act on algorithmic rent-setting, signed three days earlier, takes effect on the first day of the twelfth month after enactment, which lands on July 1, 2027. That tells you nothing definitive about this Act. It only shows the range.
Pull the enrolled text of A4523, read the final section, and put that date in the compliance calendar. Do not take it from a news summary, including this one.
How does this compare with other algorithmic pricing rules?
The sharpest contrast is with the other New Jersey law from the same week. Both target algorithmic pricing. They rest on different legal theories, which changes what a regulator has to prove.
| Instrument | Goods or services covered | Is individual personal data the trigger? | Enforcement channel |
|---|---|---|---|
| NJ Fair Price Protection Act (A4523), signed July 23, 2026 | Necessities, including groceries. Not general retail. | Yes. Personal information about the specific shopper, plus an identical product sold at a different price. | New Jersey Attorney General, consumer protection. |
| NJ FAIR Act (A3497, P.L.2026, c.43), signed July 20, 2026 | Residential rent. | No. Coordination through shared rent-setting software across competing landlords. | New Jersey Antitrust Act, Attorney General, with rulemaking authority. |
| Federal antitrust action on algorithmic rent-setting (US Department of Justice) | Residential rent. | No. Coordination and information-exchange theory. | Federal civil antitrust litigation and consent decrees. |
| Canada, Competition Bureau posture on AI-related pricing and marketing claims | Advertising and pricing representations generally. | No. A misleading representation to consumers. | Competition Bureau under the Competition Act. |
The pattern is worth naming. Almost every algorithmic pricing rule in force elsewhere is an antitrust or deception rule, so the regulator has to establish coordination between competitors or a false claim to the public. New Jersey's Act skips both. A single firm acting alone, making no representation to anyone, can breach it by pricing an individual off that individual's data. For a single-brand retailer, that is a new exposure.
What about the electronic shelf label moratorium?
The Act also imposes a one-year moratorium on new deployments of electronic shelf labels, pending a study by the New Jersey Innovation Authority. Digital shelf tags are the hardware that makes rapid in-store price changes practical, and the legislature wanted to understand them before they become ubiquitous rather than after.
For a grocery operator mid-rollout, the operative word is new. A moratorium on new deployments treats existing installations differently from planned ones, and the boundary between finishing an in-flight programme and starting a fresh one is where disputes sit.
What should a pricing team do about it?
Start with a factual question, because the legal answer follows from it. Does your pricing system ever produce a different number for two shoppers, at the same moment, for the same item, because of something it knows about one of them? Many teams do not know. Personalisation logic accretes over years, and the price surface is often assembled from a promotions engine, a segmentation model and a vendor component nobody has read end to end.
If the answer is yes, separate the mechanisms. A published loyalty discount, a coupon, a segment-wide promotion and an individually optimised price are four different things that often live in the same codebase and produce the same visible outcome. New Jersey treats them differently. Your documentation should too, written before anyone asks.
Then check the catalogue against the necessities definition, and keep the vendor question live. If a third party supplies the personalisation model, you still own the price the shopper sees.
Frequently asked questions
What is surveillance pricing under the New Jersey Fair Price Protection Act?
The Governor's signing release describes it as the use of personal information, such as online activity, location, purchasing history or other collected data, to charge different prices for identical products based on what an algorithm predicts a shopper is willing or able to pay. The trigger is data about the specific individual. Prices that move for everyone with time, inventory or demand are a different practice.
When does the Fair Price Protection Act take effect?
The effective date could not be verified from the primary source. The Governor's release of July 23, 2026 announces the signing but does not state an operative date, and the operative-date clause of A4523 was not read for this entry. Check the enrolled text of A4523 before setting a compliance deadline.
Does the Act ban loyalty programs and discounts in New Jersey?
No. The measure expressly preserves loyalty programs and ordinary discounts. What it targets is charging different prices for identical products based on an algorithmic prediction of an individual shopper's willingness or ability to pay.
What does the Act do about electronic shelf labels?
It imposes a one-year moratorium on new deployments of electronic shelf labels, pending a New Jersey Innovation Authority study.
Who enforces the New Jersey surveillance pricing ban?
The New Jersey Attorney General. There is no private antitrust-style coordination theory to prove and no federal agency in the loop. Enforcement runs through the state consumer protection channel.
Last verified: July 28, 2026