Submitted the NYT article on Maryland's ban, highlighting the framing of AI-driven price increases as a consumer harm. The story's 183 points and 141 comments suggest strong community interest in the regulatory intervention against per-customer price differentiation.
Frames the legislation as targeting 'surveillance pricing' — the practice of using customer data, purchase history, location signals, and behavioral patterns to set opaque individualized prices. The article positions Maryland as a first-mover converting the FTC's 2024 investigatory posture into enforceable state-level prohibition.
Argues that behind every dynamic price tag is a software stack involving vendors like Revionics, PROS Holdings, Bloomreach, Blue Yonder, and Eversight. Points to Kroger's partnership with Microsoft on AI-driven pricing and ESL deployment, Walmart's ESL rollout, and Amazon's algorithmic pricing at Whole Foods as evidence that this is fundamentally a technology industry issue, not just a retail one.
Draws a careful distinction between banned practices and legitimate ones: traditional promotional pricing, loyalty card discounts (uniform and opt-in), and markdown optimization on perishable goods remain legal. The law specifically targets systems using personal consumer data to charge different prices to different shoppers for the same product — the opaque, per-customer differentiation that consumers cannot see or understand.
Reports that Maryland's law converts the FTC's 2024 orders (targeting eight companies including Mastercard, JPMorgan Chase, Accenture, McKinsey, and pricing algorithm vendors) into actual enforceable prohibition at the state level. Notes explicitly that 'other states are watching,' positioning this as likely the first domino in a broader regulatory trend.
Maryland has become the first state in the US to sign legislation banning AI-driven price increases in grocery stores. The law targets what regulators and consumer advocates call "surveillance pricing" — the practice of using customer data, purchase history, location signals, and behavioral patterns to set individualized or dynamically adjusted prices on groceries.
The legislation lands in a regulatory environment that's been heating up since the FTC issued orders to eight companies in mid-2024 investigating surveillance pricing practices. Those companies included Mastercard, JPMorgan Chase, Accenture, McKinsey, Revionics (now Aptos), Bloomreach, PROS Holdings, and Marchex — a mix of data brokers, consultancies, and the actual pricing algorithm vendors that grocery chains license. Maryland's law converts the FTC's investigatory posture into actual enforceable prohibition at the state level, and other states are watching.
The bill specifically addresses systems that use personal consumer data to charge different prices to different shoppers for the same product. This is distinct from traditional promotional pricing, loyalty card discounts (which are uniform and opt-in), or markdown optimization on perishable goods. The target is the opaque, per-customer price differentiation that consumers cannot see or understand.
### The surveillance pricing ecosystem is bigger than most developers realize
Behind every dynamic price tag in a grocery store is a software stack. Kroger partnered with Microsoft on AI-driven pricing and deployed electronic shelf labels (ESL) from vendors like Pricer AB through their EDGE Shelf system. Walmart has rolled out ESL across stores. Amazon runs algorithmic pricing across Whole Foods and Amazon Fresh with prices changing frequently based on competitive intelligence. The tech vendors powering this — Revionics, PROS Holdings, Bloomreach, Blue Yonder, Eversight (now Instacart), dunnhumby, SymphonyAI — represent a multi-billion dollar B2B market built on the premise that AI-optimized pricing is strictly better than static pricing.
Maryland's law doesn't ban dynamic pricing wholesale — it bans the use of personal consumer data to set discriminatory prices. That's an important distinction. A grocery chain can still lower prices on expiring produce (markdown optimization) or adjust prices based on competitive pressure. What it cannot do is charge Customer A more than Customer B for the same box of cereal because Customer A's purchase history suggests higher willingness to pay.
### The Wendy's precedent
The political momentum for this legislation traces back to February 2024, when Wendy's announced plans for "surge pricing" on menu items — dynamically adjusting prices based on demand, time of day, and other signals. The backlash was immediate and severe. Wendy's walked it back within days, rebranding the initiative as offering "dynamic discounts." But the damage was done: the phrase "surge pricing for food" entered the political lexicon, and legislators took notice.
Groceries carry even more political weight than fast food. Food prices are one of the most visceral economic indicators for voters. When the FTC published its January 2025 interim report on surveillance pricing, concluding that these practices are "already prevalent" and "opaque to consumers," it gave state legislators both the evidence and the vocabulary to act. Maryland moved first, but it won't be last — California, Minnesota, Connecticut, and New York have all had legislators circulating similar proposals.
### The compliance iceberg
For engineering teams building or maintaining retail pricing systems, this creates a familiar but painful compliance challenge: a state-by-state patchwork. If you've worked in ad tech (CCPA, state privacy laws), fintech (state money transmitter licenses), or healthcare (state telehealth regulations), you know the pattern. A single state passes a law. Others follow with slightly different definitions and requirements. Suddenly your pricing engine needs jurisdiction-aware logic, audit trails proving that price differentiation isn't based on protected personal data, and the ability to demonstrate compliance to regulators who may not understand your ML pipeline.
The technical challenge is nontrivial. Modern pricing optimization systems don't have a clean separation between "personal data signals" and "market signals" — they're trained on datasets where individual purchase patterns are entangled with aggregate demand curves. A model that learned pricing patterns from loyalty card data doesn't have a toggle you can flip to make it "not use personal data." Compliance may require retraining models, not just adding a filter.
If you work on retail pricing, recommendation, or personalization systems, here's what to do now:
Audit your data inputs. Map every signal your pricing model uses back to its source. Can each signal be classified as "personal" (tied to an identifiable consumer) vs. "aggregate" (market-level demand, competitor pricing, seasonality)? If you can't make that classification cleanly, you have a compliance gap. Build this audit capability now, before you're doing it under regulatory pressure with a deadline.
Design for jurisdiction awareness. Your pricing service needs to know where a transaction is happening and apply jurisdiction-specific rules. This is table stakes for any system that will operate across US states in 2026-2027. If your architecture assumes uniform pricing logic nationwide, start planning the refactor.
Prepare for explainability requirements. Regulators will eventually ask: "Why did this customer see this price?" If your answer is "the model determined it," that won't satisfy anyone. Invest in pricing explainability tooling — not just for compliance, but because the retailers licensing your software will demand it as a procurement requirement.
Watch the markdown optimization carve-out. The law appears to preserve the ability to dynamically price perishables to reduce food waste. This is the use case where dynamic pricing has the strongest public-interest argument (an estimated 20-30% waste reduction). If you're building pricing systems, lean into this framing. The grocery chains that survive this regulatory shift will be the ones that can credibly say: "Our AI reduces prices and waste, it doesn't gouge individuals."
Maryland's law is a first-mover signal, not an endpoint. The FTC's surveillance pricing report gave every state AG and legislator a roadmap. The grocery industry's lobbying power is significant, but "your grocery store is using AI to charge you more" is a message that resonates across every demographic. For developers and engineering leaders in retail tech, the strategic question isn't whether algorithmic pricing regulation is coming — it's whether your systems are architecturally ready for a world where the rules vary by state and evolve annually. The teams that treat this as a data architecture problem today will ship compliant features. The teams that treat it as a legal problem to defer will be rewriting pipelines under deadline pressure in 18 months.
Massachusetts has had fair pricing laws for grocery stores for years that I suspect already de-facto ban "dynamic pricing". It requires grocery stores to ring up the item at the lowest marked or advertised price, or the item is free. It also requires all items to be marked (or have scanner
Grocery stores have smaller margins and more options compared to pretty much any industry, yet politicians seem to think they are the cause of all of our ills.
Why grocery stores only? It’s also unclear how this will change anything - don’t the grocery stores in richer areas already charge more? I’ve noticed Whole Foods prices are not the same across all stores even in the same state.
Wait, isn't this prohibited already? Some of it may be a gray zone, but a good portion of it is already downright illegal in many countries, and the rest is extremely unethical.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
https://archive.is/2026.05.01-224445/https://www.nytimes.com...