Massachusetts puts sportsbook AI under the microscope

• Massachusetts Gaming Commission to review AI use across licensed sportsbooks.
• Review follows a New York Times investigation into DraftKings’ promotional targeting.
• Regulator has made no finding of wrongdoing and will engage DraftKings first.

The Massachusetts Gaming Commission (MGC) is reviewing how its licensed sports betting operators use artificial intelligence, starting with DraftKings.

Chair Jordan Maynard announced the move at a public meeting on September 24, following a New York Times investigation into the operator’s promotional practices.

Inside the investigation

The Times report, published on September 19, alleged that DraftKings directed staff to build a machine learning model. Its purpose was to identify customers most likely to respond to promotions and then keep losing money.

The Times said its reporting drew on interviews with more than 40 former employees. It also cited internal company communications and customer betting records.

According to the report, work on the model began in 2023. The system reportedly weighed betting frequency, account balances and historical losses. It then scored customers on the revenue promotions were expected to generate.

Coverage of the investigation also described an internal elasticity metric. It was used to estimate how much more a customer might wager after receiving an offer.

Promotions sit at the core of the strategy. According to the Times, DraftKings spends hundreds of millions of dollars each year on free bets, profit boosts and deposit bonuses. The model was meant to show which offers delivered the best return.

Jayden Butts, a former DraftKings data analyst, evaluated the model in a 2023 assignment focused on casino users. He told the Times the logic behind it disturbed him.

“The best investment would be a problem gambler.”

DraftKings disputes the characterization. The Boston-based operator told the Times it rejects any suggestion that its marketing is unfair or improperly targets customers. It has also said its targeting focuses on customers showing sustained engagement, not on losses.

The company told the news outlet that it does not send promotions to customers already flagged for risk or to those who have excluded themselves. It said it also uses large language models to scan customer messages for signs of distress.

Former employees said that those measures were mostly reactive. A separate internal project, led by former data scientist Nestor Hernandez, aimed to flag at-risk customers days or weeks earlier. The goal was to predict problems rather than react to them.

Facts before action

Maynard said he had asked Executive Director Dean Serpa and commission staff to engage with DraftKings. The aim is to understand the specifics of what was reported. The regulator stressed it has not determined that the operator committed any wrongdoing.

Jordan Maynard, Chair of the Massachusetts Gaming Commission, said:

“The commission is aware of a report from The New York Times regarding the use of AI and machine learning by an operator. These AI technologies are evolving rapidly across our society and we share the concerns over their application.”

The scope extends beyond one company. Staff will also examine how other licensed sports betting operators deploy AI across promotions, customer acquisition and responsible gaming.

Commissioner Paul Brodeur described some of the issues raised as troubling. He stressed the need to establish the facts first. Brodeur added that the risk of AI exploiting customers applies to any customer-facing business, not just gambling.

Rules already written

Massachusetts is not starting from a blank page. Brodeur noted it may be the only state whose regulations reference AI at all.

Its data privacy rules bar sports wagering operators from promoting wagers or offers through algorithms, machine learning or AI known or reasonably expected to make a platform more addictive.

Separate advertising rules prohibit operators from using the characteristics of at-risk or problem bettors to target them with ads. Operators must also report to the commission at least every six months on trends in their patron data and efforts to curb addictive behavior.

The MGC has tracked the issue for some time. It funded a research study by the UNLV International Gaming Institute into AI use in promotions, player acquisition, responsible gaming tools and player health. Maynard also pointed to the commission’s own AI task force.

The review lands at a sensitive moment for DraftKings. A California class action, filed on August 13 in federal court, already alleges its Predictions product amounts to unlicensed sports betting.

The Massachusetts process also turns a long-running industry debate over AI in CRM into a live compliance question. The inquiry could test how the MGC applies its existing prohibition in practice.

A stricter reading could push operators to document how promotional models account for markers of gambling risk. Any finding is also likely to draw attention from regulators in other states.