How AI could automate the poor tax

Imagine two people shopping online for the exact same item at the exact same time, yet one receives a higher price because an algorithm predicts they are willing to pay it.

That possibility sits at the center of growing concerns around surveillance pricing and AI-driven personalized pricing systems — concerns serious enough that Maryland became the first state, and Connecticut the second, to enact laws restricting certain forms of surveillance pricing. That momentum has continued in New York and New Jersey, where lawmakers have passed similar legislation awaiting their governors’ signatures. 

Although these efforts share a common goal, they differ in scope. Maryland and New Jersey focus primarily on grocery stores. Connecticut extends its restrictions more broadly to retail transactions, and New York’s proposal would apply across industries.

Together, these laws reflect a growing concern that AI could shift pricing from supply and demand to individualized predictions. Instead of asking, “What is this product worth?” AI asks, “What is this consumer willing to pay?”

For many Black Americans — whose “households have roughly 15 cents for every one dollar in wealth that White households have” — and other low-income Americans, that concern feels deeply familiar.

Low-income communities have long recognized what is often called the poor tax. Financially vulnerable communities frequently pay more over time — through predatory lending, overdraft fees, subprime financial products, higher insurance costs and limited access to affordable goods and services. The concern is now that AI could automate and scale those inequities, making them harder to detect.

These systems would not need to know a consumer’s income directly. Modern AI can infer purchasing power and consumer behavior from purchase histories, browsing activity, location data, loyalty program participation, device information, shopping........

© The Hill