The Massachusetts Gaming Commission wants every licensed sports wagering operator to explain how it uses artificial intelligence and machine learning in marketing, promotions and customer engagement. Written responses are due 3 November.
The request came from Executive Director Dean Serpa. Trade press dates his letter to 9 October, and the commission published its notice on 14 October. The notice says operators "have been requested" to detail their AI and machine learning use, and it sets the deadline. It does not describe the response as compulsory.
What the commission is asking
The letter goes to the seven licensed mobile operators. According to trade coverage of its contents, it covers marketing, promotions, customer engagement and how operators assess potentially harmful gambling. Three questions carry most of the weight:
- Which systems: each model, tool or algorithm used for marketing, promotions or customer engagement, and what it is for.
- Who built it: whether each system was built in-house or supplied by a third party.
- Whose decisions: whether the system makes decisions about individual customers, groups of customers, or both.
Trade coverage also reports a question on whether a patron's losses or rising activity make that patron more likely to receive a promotion. The letter asks about human oversight of these systems too.
Why now
The review follows a directive from the commission's chair, Jordan Maynard, issued on 24 September. Trade coverage ties it to a New York Times report that DraftKings used machine learning to find customers likely to keep gambling and losing after receiving offers. The commission has not found that DraftKings did anything wrong.
The same question is moving elsewhere. Louisiana's board deferred a full renewal of DraftKings' licence while it examines the company's AI use ([[louisiana-draftkings-conditional-extension]]). Maine's gaming regulators are monitoring AI deployment in sportsbooks ([[maine-sportsbook-ai-monitoring-regulations]]). Three regulators are now asking the same type of question about AI in sportsbook marketing and pricing.
The real line is the individual offer
The most important distinction in the letter is between decisions about groups and decisions about individuals. A group-level promotion, such as a boost for a fixture or a loyalty tier, is a marketing choice. An engine that decides, for each bettor, whether a loss should trigger a bigger or more frequent offer is a product decision with a direct effect on play.
The question about losses is the sharpest one because it asks what the model treats as a signal. A system that reads losing as a reason to offer more is not necessarily illegal. It is, however, the mechanism regulators will test against harm standards. Operators that cannot explain which inputs drive an individual offer will struggle to show they are not using losses as a trigger.
The vendor gap
Asking whether a system is in-house or third-party is a practical question. Sportsbooks often license promotional and risk engines from suppliers, and the operator may hold the configuration but not the underlying logic. Under that arrangement, an operator can know what an offer is supposed to do without knowing how the model reaches it.
The commission's question forces a split. Operators will have to say which tools they built, which they bought, and who can change the rules inside each. Vendor contracts that treat the model as a black box will make the answer hard to give.
Human oversight is the next test
The reference to human oversight suggests the commission wants a documented control, not a description of the model. The questions an operator should be able to answer include who can override an automated offer, which thresholds trigger a human review, and how often those reviews happen. Those are operational facts. If they are not written down, the oversight claim is weak.
This is also where responsible gambling tooling meets promotions. AI in risk management is already a standard part of operator controls ([[ai-responsible-gambling-integration]]). Regulators are now asking whether the same models that flag harm also drive the offers that create it.
What the November responses will show
The responses are the commission's evidence base for its next move, whatever that is. The first useful signal will be whether operators can describe their models at all, and whether their answers match the offers their customers receive. A response that says the system is proprietary and nothing more will invite a second, more formal request.
The letter does not, by itself, change any rule. It does make the offer engine a subject of regulatory inquiry, and that is a shift that will outlast this batch of answers. Operators that treat the letter as a marketing-compliance exercise have misread which part of the business is being examined.