When two shoppers browse for the same product at the same time, they might see different prices—one higher because an algorithm decides they can afford it. That scenario is driving a wave of state legislation targeting what critics call surveillance pricing, a practice that could automate and scale the financial burdens long borne by low-income Americans.
Maryland became the first state to enact such a law, followed by Connecticut. New York and New Jersey have passed similar measures awaiting their governors' signatures. While all aim to curb AI-driven personalized pricing, their scope varies: Maryland and New Jersey focus on groceries, Connecticut extends to retail broadly, and New York's proposal would cover all industries.
These laws reflect a fundamental shift in pricing logic. Instead of setting prices based on supply and demand, AI systems ask, “What is this consumer willing to pay?” That shift threatens to entrench economic disparities, particularly for Black Americans—who hold about 15 cents of wealth for every dollar held by white households—and other low-income groups.
Low-income communities have long recognized the “poor tax”: higher costs for predatory loans, overdraft fees, subprime financial products, and limited access to affordable goods. AI could amplify these inequities, making them harder to detect. These systems don't need direct income data; they infer purchasing power from purchase histories, browsing behavior, location, loyalty programs, and device info. Individually trivial, these signals build detailed profiles that predict how much a consumer will pay.
Unlike traditional discrimination, algorithmic pricing operates invisibly. Consumers see only the final price, unaware of the data collected or whether they've been categorized differently. That opacity is central to the debate.
Americans accept dynamic pricing for airlines, hotels, and ride-sharing, where prices fluctuate with market conditions. Surveillance pricing differs by targeting individual consumers, not general demand. Maryland's law distinguishes between the two, defining dynamic pricing as within-day variation based on demand, while surveillance pricing uses personal data to set personalized prices.
At a time of rising food costs and housing instability, the prospect that algorithms could identify the most financially constrained shoppers and charge them more raises profound fairness questions. The concern is not hypothetical: Walmart faced backlash after discussing electronic shelf labels that could enable dynamic grocery pricing, and Delta Air Lines drew criticism for exploring AI that gauges individual willingness to pay. Both companies later clarified their policies, but the incidents underscored public unease.
Once such systems become embedded, regulation grows harder. Maryland acknowledges enforcing its law may require technical expertise beyond traditional consumer protection agencies. Regulators might need to probe machine learning models, behavioral analytics, and third-party data brokers—systems operating at a pace consumers can't access or understand.
This challenge reflects a broader misunderstanding of AI, often seen as an objective tool rather than a product of the economic systems it serves. If the digital economy rewards behavioral prediction and profit maximization, AI will optimize for those priorities. States are now trying to intervene before that optimization becomes irreversible, but the technical and political hurdles are steep.
