New York proposal draws a line between player protection and AI promotion

In brief

• The New York State Gaming Commission advanced a revised proposal restricting how mobile sports betting licensees use AI, algorithmic profiling and other automated systems for promotions and wagering advice.
• The regulator also advanced a proposal requiring licensees to identify patrons who may be at risk of problem-gambling behavior.
• The proposals are not final rules, and the revised AI text still faces public comment.

CDC Gaming reported the commission's action on 11 October. The development matters beyond one jurisdiction because the same customer data can power both protective intervention and commercial personalization.

What moved forward

The revised AI rules will receive additional public comment before returning for discussion and a possible vote. No date has been set for a final vote.

CDC Gaming says the proposal covers automated use of customer data for personalized promotions, bonuses, messages, product recommendations and wagering suggestions. The exact obligations should be checked against the final regulatory text when it is published.

Earlier official draft shows the policy direction

The March official draft considered restricting all licensees or franchisees, not only sports wagering, from AI-personalized promotions and wager suggestions. Governor Kathy Hochul's office also described a proposal for risk indicators, a responsible-gaming lead and staged intervention. That page records the earlier draft, not the final wording of the revised proposal.

Keep protection separate from promotion

Slotgen's practical takeaway is architectural: customer-level recommendations and responsible-gambling detection should not share an unexplained decision path. Record the purpose of each model, its input window and version, the reason for each output and the human owner who can review it.

A protection flag should be able to block a promotional recommendation without erasing the audit trail. Consent and marketing preferences should be explicit inputs, while staff should be able to see whether a message came from a fixed rule, a model or a human decision.

What is not yet known

The sources do not establish the final scope, implementation date, technical tests or enforcement process. They also do not show that any particular AI design is compliant. Operators should wait for final text and obtain jurisdiction-specific legal advice before changing regulated workflows.