In 2026 New Jersey and New York moved from asking businesses to disclose algorithmic pricing to banning prices set from a customer's personal data, which turns the data feeding pricing engines and AI agents into a compliance boundary that security and identity teams have to enforce. This guide sets out what New Jersey's Fair Price Protection Act and FAIR Act and New York's Algorithmic Pricing Disclosure Act and One Fair Price Act say, how they are enforced, and which data, access and logging controls let an organisation prove its prices are not built from personal data.
It is written for privacy, security and identity teams at retailers, grocery and delivery platforms, marketplaces, property managers and the software vendors that supply their pricing systems. The wider security of commerce agents, including pricing agent integrity and competition law, is covered in AI Agents in Retail and Ecommerce; this page focuses on the pricing laws themselves.
Status as of 17 Nov 2026. This guide is a practitioner summary and is not legal advice. Some figures differ between published summaries; where they do, we say so. Confirm the enacted text and current status with counsel before relying on it.
Key takeaways
- New York's Algorithmic Pricing Disclosure Act (General Business Law section 349-a) took effect on 10 November 2025. It requires the notice "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA" and carries civil penalties of up to $1,000 per violation.
- New York's legislature passed the One Fair Price Act (A9349B / S8623B) on 4 June 2026. It would replace disclosure with a ban on "surveillance pricing", with penalties of up to $5,000 for a first violation and $20,000 for each later one. At the time of writing it awaited action by the Governor, who has until 31 December 2026.
- New Jersey's Fair Price Protection Act was signed on 23 July 2026 and bans personal-data pricing for groceries and other foodstuffs from 1 August 2027. It is reported as the first such law with a private right of action, and it pauses new electronic shelf labels for a year from 1 February 2027.
- New Jersey's FAIR Act, signed on 20 July 2026 and effective 1 July 2027, bans landlords and third parties from using algorithmic devices to coordinate rents.
- Loyalty programmes and broad discounts survive under both states' laws, but only within defined conditions. Dynamic pricing that does not use personal data remains lawful, with disclosure duties in New York's bill.
- Compliance depends on knowing which data reaches the pricing model, which identities can change it and whether you can show that history to a regulator.
Personalised pricing laws at a glance
| Law | Status as of 17 Nov 2026 | What it does | Who enforces |
|---|---|---|---|
| New York Algorithmic Pricing Disclosure Act (GBL 349-a) | In force since 10 November 2025 | Requires a disclosure when an algorithm sets a price using the consumer's personal data | Attorney General; civil penalties up to $1,000 per violation |
| New York One Fair Price Act (A9349B / S8623B) | Passed both houses 4 June 2026; awaiting Governor; effective 180 days after becoming law | Bans surveillance pricing and the collection, use, sale or sharing of personal data to facilitate it | Attorney General; up to $5,000 first violation, $20,000 each later one |
| New Jersey Fair Price Protection Act | Signed 23 July 2026; main provisions effective 1 August 2027 | Bans personal-data pricing for groceries and other foodstuffs; one-year pause on new electronic shelf labels | Attorney General under the Consumer Fraud Act; private right of action |
| New Jersey FAIR Act | Signed 20 July 2026; effective 1 July 2027 | Bans algorithmic coordination of rents and sharing of sensitive market data | Treated as a violation of the New Jersey Antitrust Act |
Other states are moving too. One summary of the New York bill notes that Maryland has prohibited dynamic pricing for food retailers and third-party delivery providers and that Connecticut has enacted a similar prohibition alongside a personalised pricing disclosure rule, with California also considering proposals. Track these through the US State AI Laws Tracker.
Scope
New York
The current law, GBL 349-a, applies when a price is set by an algorithm using a consumer's personal data. It does not ban the practice; it requires the disclosure above.
The One Fair Price Act, as passed, reaches any entity and its service providers. It defines surveillance pricing as pricing set fully or partly by an algorithm that uses personal data to offer different prices to different customers for the same goods or services, excluding bona fide custom discounts. Personal data means data that identifies, or could reasonably be linked to, a specific consumer or device, with limited exceptions for certain location data. "Algorithm" is defined broadly as applying one or more sets of rules to generate outputs from inputs, which reaches rules engines as well as machine learning and AI agents.
The bill excludes entities subject to insurance law, financial institutions and pricing or credit decisions based on creditworthiness from consumer report data, and pricing required or expressly authorised by federal or state law. Service providers are excluded unless they show reckless disregard of the entity's prohibited conduct, which matters for pricing software vendors.
New Jersey
The Fair Price Protection Act applies to the sale of "groceries and other foodstuffs", a broad category that published analysis says includes dairy, meat and deli, produce, seafood, beverages, snacks, baked goods, paper products, household cleaning items, health and beauty products, frozen foods and pet supplies. The prohibition applies to "any person" and appears to reach grocery retailers and third-party delivery platforms operating in New Jersey.
It defines surveillance pricing as a price "determined, adjusted, optimized or recommended by an algorithm or automated system" based on personal data, producing price variation for individual consumers or groups. The words "recommended" and "automated system" bring AI pricing assistants and agents into scope even where a person approves the final price.
The FAIR Act covers landlords and third-party coordinators using an "algorithmic device", meaning any device that uses algorithms to process data on rental prices, occupancy levels or material lease terms. A spreadsheet that works without AI and needs human analysis, and a database that only queries unprocessed stored data, are excluded.
Key obligations
New York: disclose now, stop if the ban is signed
- Today. Where an algorithm sets a price using personal data, show "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA".
- If the One Fair Price Act becomes law. Do not set or adjust prices using surveillance pricing, and do not advertise it. Do not collect, use, sell, share or disclose personal data to facilitate it. The bill would remove the disclosure route and replace it with the ban.
- Discounts. Bona fide custom discounts, such as loyalty programmes, voluntary self-identification as a veteran or senior, or prior purchase history with the same entity, must be disclosed with their eligibility criteria and offered uniformly to eligible consumers.
- Dynamic pricing. Algorithms that vary prices for non-personal reasons stay lawful. Published analysis says businesses using them must clearly disclose that the tactic is used, how often prices may change and the conditions that affect pricing, and that these duties apply to algorithms changing prices more than once a day.
New Jersey: groceries
- Do not use surveillance pricing, or any other strategy that sets or varies grocery prices in whole or in part on personal data. Doing so violates the New Jersey Consumer Fraud Act.
- Prices may differ according to the reasonable cost of serving different consumers, but a given price may not change more than once in 24 hours.
- Discounts for broadly defined groups, such as teachers or senior citizens, are allowed if eligibility is publicly and conspicuously disclosed and applied uniformly.
- Loyalty discounts, including those based on purchase history, are allowed only if enrolment is voluntary and opt-in, all members receive uniform pricing benefits, participants get clear disclosure of benefits and data practices, and the programme publicly discloses its benefits and data practices.
- Electronic shelf labels connected to surveillance pricing technology are prohibited, and new electronic shelf labels are paused for one year from 1 February 2027. Existing labels may be used, repaired and replaced. The New Jersey Innovation Authority is studying their effects.
New Jersey: rents
- Landlords and third-party coordinators must not use algorithmic devices to coordinate rental prices.
- They must not share sensitive market data, such as rental prices, occupancy levels or material lease terms, where the sharing facilitates coordination.
Enforcement and penalties
| Law | Enforcement route | Penalties and remedies |
|---|---|---|
| NY GBL 349-a (current) | Attorney General | Up to $1,000 per violation |
| NY One Fair Price Act (as passed) | Attorney General; injunction on five days' notice; rulemaking power | Up to $5,000 first violation, $20,000 each later one; restitution and damages; one summary adds that the penalty may instead equal profits from the violation if greater. No private right of action in the passed version. |
| NJ Fair Price Protection Act | Attorney General under the Consumer Fraud Act; private right of action, class actions permitted | Consumers: actual damages, treble damages for wilful violations. Attorney General: one law firm summary reports the greater of actual damages or $50,000 per violation, while another reports fines of up to $20,000; check the enacted text. |
| NJ FAIR Act | New Jersey Antitrust Act; Attorney General complaint database | Civil penalties, injunctive relief and treble damages |
The private right of action in New Jersey changes the risk profile. A consumer who shows they were charged more than other customers for identical groceries because of their personal data can sue, and class actions are permitted. Discovery in such a case will ask what data the pricing system used, which is a data lineage question before it is a legal one.
What it means for security and identity teams
These laws are drafted as consumer protection rules, but compliance rests on controls that security and identity teams already run.
Data lineage into pricing
The test in both states is whether personal data was used to set, adjust, optimise or recommend a price. To answer that, you need to know every feature that reaches a pricing model or rules engine. Browsing history, precise location, device information, purchase history and inferred demographics are the examples given in published analysis. A customer data platform feed, an identity graph join or a session cookie passed to a pricing service can each bring personal data in. Treat the pricing system's inputs as a governed data flow with a documented allow list.
Separating loyalty from pricing
Loyalty discounts are lawful only under conditions. The simplest technical design keeps the base price engine free of personal data and applies published loyalty discounts afterwards, from a separate service that holds only programme membership. That separation is easier to prove than a single model with a compliance flag.
Pricing agents are non-human identities
Where an AI agent or automated service recommends or sets prices, it is an identity with access to data stores and the ability to write prices. Give it its own credential, scope its read access to non-personal sources, and deny it access to customer profile, clickstream and location stores (NIST SP 800-53 AC-6 Least Privilege). Review its entitlements like any privileged service account (AC-2 Account Management). An agent that can query a customer profile while computing a price is the easiest way to breach these laws without anyone deciding to.
Change control and audit trails
Log model versions, feature sets, rule changes and who approved them, and keep price change history long enough to answer a complaint. New Jersey's once-in-24-hours limit on cost-based variation and New York's disclosure duty for algorithms that change prices more than once a day both need price change logs to demonstrate compliance (AU-6 Audit Record Review, Analysis, and Reporting).
Vendors and service providers
New York's bill excludes service providers unless they show reckless disregard, which gives vendors a reason to ask customers what data they feed in and gives customers a reason to ask vendors what their models use. Write data use limits into pricing software contracts and confirm that vendors do not enrich your feeds with third-party personal data. The US State Data Broker Laws guide is relevant where enrichment data comes from brokers.
Interaction with privacy law
Personal data used for pricing is also processing under state privacy laws such as California's. Profiling and automated decision rules may apply alongside the pricing bans; see the CCPA and CPRA guide and the FTC Act Section 5 Data Security Guide for the federal unfairness angle.
Compliance steps
- Inventory pricing systems. List every engine, model, rules service, agent and third-party tool that sets, adjusts, optimises or recommends prices, including electronic shelf label platforms and delivery partner integrations.
- Map inputs. For each system, document every input feature and data source, and mark any that identify or could reasonably be linked to a consumer or device.
- Remove or isolate personal data. Cut personal data from pricing inputs in covered markets, or move personalisation into a separate, disclosed loyalty or discount layer that meets each state's conditions.
- Apply New York's disclosure now. Where personalised algorithmic pricing still runs for New York consumers, show the GBL 349-a notice until the ban takes effect, then stop the practice.
- Restrict access. Give pricing agents and services dedicated identities with no access to customer profile, clickstream or location stores, and alert on new grants.
- Log and retain. Keep model, feature and price change history with approver identity, and test that you can reconstruct why a given price was shown.
- Review loyalty programmes. Confirm opt-in enrolment, uniform member benefits and public disclosure of benefits and data practices for New Jersey.
- Plan for electronic shelf labels. Freeze new deployments in New Jersey from 1 February 2027 and confirm existing labels are not connected to personal-data pricing.
- Review rent tools. Property businesses should assess revenue management software against the FAIR Act's algorithmic device definition before 1 July 2027.
- Update contracts. Add data use restrictions and audit rights to pricing vendor agreements.
Common failure patterns
- Assuming "dynamic" means "safe". Dynamic pricing is lawful only while personal data stays out of it; a single feature such as device type or precise location can change that.
- Personal data through the back door. Segment IDs, look-alike audiences or third-party enrichment reach the model without anyone classing them as personal data.
- Agents with broad read access. A pricing agent given the same data access as a marketing agent can personalise prices as a side effect of optimisation.
- Loyalty programmes that are not uniform. Member prices that vary by individual behaviour fail New Jersey's uniform benefit condition.
- No reconstructable history. Without logs of inputs and model versions, an organisation cannot rebut a private claim or an Attorney General inquiry.
- Treating the vendor as responsible. The deploying business sets the price; a vendor model that uses personal data creates the deployer's exposure.
Practitioner checklist
- Pricing system inventory complete, including agents, electronic shelf label platforms and partner tools.
- Input feature register for each system with personal data flagged.
- Personal data removed from pricing inputs for New Jersey groceries before 1 August 2027.
- GBL 349-a disclosure shown wherever personalised algorithmic pricing reaches New York consumers.
- Status of New York's One Fair Price Act checked, with a plan for the 180-day lead time.
- Pricing agents and services run under dedicated, least-privilege identities.
- Price, model and rule change logs retained with approver identity.
- Loyalty and group discount terms reviewed against each state's conditions.
- Pricing vendor contracts include data use limits and audit rights.
- Rent revenue management tools assessed against the FAIR Act.
Standards and references
- New York State Senate: A9349 Prohibits the use of surveillance pricing
- Wilson Sonsini: New York Legislature passes ban on personalized pricing
- Skadden: New Jersey enacts first-in-the-nation surveillance pricing law with a private right of action
- Crowell: New Jersey takes aim at algorithmic and surveillance pricing
- NIST: AI 100-1 Artificial Intelligence Risk Management Framework (AI RMF 1.0)
- ISO: ISO/IEC 42001:2023 Artificial intelligence management system
- OWASP: LLM06:2025 Excessive Agency
Related NHI Mgmt Group resources: AI Agents in Retail and Ecommerce · US State AI Laws Tracker · US State Data Broker Laws · CCPA and CPRA