Massachusetts regulators demand AI transparency from sportsbooks

Paul Brodeur said it plainly last month: Massachusetts is one of the only states in the country to reference artificial intelligence in its gaming regulations. On Thursday, the Massachusetts Gaming Commission turned that distinction into a 70-question document and sent it to every licensed sports betting operator in the state.

The review began after a New York Times report alleged that DraftKings used machine learning to identify losing customers and target them with promotions designed to keep them betting. DraftKings denied wrongdoing. The MGC's response was to widen the lens — not DraftKings specifically, but every operator it licenses, and not just what they do, but how they decide.

The commission's letter sets its own definition of what counts as artificial intelligence, and it is deliberately broad. Any automated tool that ranks, classifies, or makes predictions about patrons falls within scope, regardless of whether the operator calls it personalization, optimization, or risk assessment. That definitional move matters: it closes the labeling escape route before operators can take it.

The questions are organized into eight sections. The first two establish baseline inventory — what AI systems exist, what data they consume. The remaining six dig into the mechanics of promotion delivery: who receives what, on what trigger, and whether a patron's losing history influences any of those decisions. Operators that answer no to that last question are also required to explain the controls that prevent it.

The final section addresses human oversight — specifically, at what point a person reviews a model's recommendation and whether that person has authority to override it. The MGC also asks whether employees receive an explanation of why a model flagged a particular patron for a particular promotion. These are not rhetorical questions. They test whether an operator can reconstruct its own decision chain, patron by patron, promotion by promotion.

I have seen this structure before in SEC enforcement contexts: regulators who already suspect the answer design questions that force the respondent to either admit the mechanism or document the controls that supposedly prevent it. Both answers are useful. The MGC is not running an academic survey.

Participation is listed as voluntary, and responses will be kept confidential. The voluntary framing is standard for an information-gathering phase that precedes enforcement authority. MGC Chair Jordan Maynard said last month that further action would be determined "when and if appropriate." The commission has set a November 3 deadline for responses.

Michigan's Gaming Control Board separately confirmed it is reviewing the issues raised in the Times report, though it stopped short of committing to a formal DraftKings-specific inquiry. Michigan's gaming statutes contain no provision addressing AI.

The standard that applies here is the one Massachusetts already put in regulation: AI that is "known or reasonably expected" to make a gaming platform more addictive is prohibited. The November 3 responses will either establish that operators can account for their systems against that standard, or they will establish that they cannot.