State bans put data-driven pricing under new scrutiny

Maryland and Connecticut have become the first states to prohibit certain uses of personal data in food pricing, turning long-running concern over algorithmic price discrimination into law. Ravi Dhar and Jon Iwata of the Yale School of Management argue in a Fortune commentary that the same trust questions will grow sharper as AI agents begin acting for consumers.

Dhar is a management and marketing professor and director of the Yale Center for Customer Insights. Iwata is a lecturer at Yale School of Management and executive director of Yale’s Program on Stakeholder Innovation and Management, after previously serving as IBM’s senior vice president and chief brand officer.

According to Dhar and Iwata, state efforts to restrict “surveillance pricing” failed to produce a ban during 2025 despite multiple proposals. Maryland changed that in April by barring food retailers and delivery services from using consumers’ personal data to set prices, they wrote. Connecticut followed in June, while California and New York are weighing similar measures.

The authors describe surveillance pricing as a test of how companies use data, algorithms and AI. They draw a distinction between prices that reflect market conditions and prices that reflect a company’s assessment of an individual customer’s willingness or need to pay.

In their example, riders may accept that an Uber trip from Midtown Manhattan to Newark Airport costs more during bad weather or heavy demand than during a slower period. They argue consumers are more likely to object if two riders in the same place at the same time receive different prices because of their data profiles, buying histories, devices or inferred willingness to pay.

The Federal Trade Commission’s 2025 study showed that algorithms using personal data can infer when consumers have fewer choices, stronger urgency or higher willingness to pay, Dhar and Iwata wrote. They said similar tools can affect workers, such as when a platform predicts a driver would accept a lower payout because she is close to an earnings target or unlikely to move to another app.

Dhar and Iwata argue that companies risk moving from efficient market pricing toward extraction when they use personal vulnerabilities to set prices or offers. They said that approach can weaken trust, increase worker dissatisfaction and draw more regulation.

The authors say AI agents could make the issue more difficult because they may collect a fuller picture of users’ lives than search engines or individual apps. As people assign agents to book travel, order goods or manage household tasks, those systems may learn preferences, routines and needs that users do not state directly.

They cite Bain & Company’s estimate that AI agents could influence $300 billion to $500 billion in U.S. commerce by 2030. Dhar and Iwata say that influence could support useful personalization, such as finding lower fares, prompting medication refills or filtering irrelevant offers.

The same capabilities could also be used to charge each person closer to the maximum they are likely to pay, contact them at vulnerable moments or withhold better options when they are likely to accept weaker ones, according to the authors. They frame the central issue as whether an agent serves the user, the platform that built it or advertisers and sellers seeking recommendations.

Dhar and Iwata point to Yale’s Program on Stakeholder Innovation and Management, where they say interviews with more than 200 CEOs found that long-term shareholder value is best built by growing trust with customers, workers, suppliers and communities. They argue that surveillance pricing is an early test of how companies will govern more powerful AI tools.

This story draws on original reporting from Fortune.