Massachusetts to Review How Sportsbooks Use AI After the DraftKings Report

The Massachusetts Gaming Commission has told its staff to question DraftKings over the New York Times investigation and to survey how every licensed sportsbook uses AI and machine learning. It has not found wrongdoing, and any rule-making would come later.

• The Massachusetts Gaming Commission (MGC) has asked Executive Director Dean Serpa and staff to engage with DraftKings on the use of AI and machine learning described in a New York Times investigation, Chairman Jordan Maynard said on 24 September
• Maynard also asked staff to examine how all licensed operators use these technologies, and said any further action or policy would be for the commission to decide "when or if appropriate"
• The Times reported that DraftKings built a model in 2023 that scored online casino customers by expected losses per promotion, and shelved a separate model to flag customers approaching harm
• DraftKings, which is headquartered in Boston, told the Times that it rejects any implication that its marketing improperly targets customers
• The review will draw on an AI task force the MGC set up after a study it commissioned from UNLV found a "governance gap" between the technology and existing rules

The Regulator in DraftKings' Home State Asks for the Facts

Jordan Maynard, chairman of the Massachusetts Gaming Commission, announced at a public meeting on Thursday 24 September that he had asked Executive Director Dean Serpa and staff to engage with DraftKings "to understand the specifics" of the New York Times report, and to see "how all operators are using these technologies today", according to CDC Gaming and Covers. "These technologies are evolving rapidly across our society and we share the concern over their application," Maynard said. "Any further action or policy making can then be determined by the Commission when or if appropriate."

The Times investigation, published on 19 September, reported that DraftKings built a machine-learning model in 2023 that gave each online casino customer an "elasticity" score, with a higher score meaning more expected loss for every promotion sent, and that a separate model intended to flag customers heading towards harm was shut down in early 2025. It drew on internal memos, Slack messages, betting records and interviews with more than 40 former employees. DraftKings told the paper that promotions go to customers with "sustained, engaged use", that the 2023 test was "preliminary and inconclusive", and that it decided against predictive harm tools because the approach was not "evidence-based".

The commission has not made any finding against DraftKings. Commissioner Paul Brodeur said the regulator needed "to find out the facts on the ground, with no disrespect to The New York Times", and said the questions apply to "any retail experience, anything that is customer-facing", according to Covers.

A Study, a Task Force and Rules That Already Mention AI

Maynard said the commission had been following AI in gambling before the report. In November 2025 it published the AI and Player Risk Identification and Response report, commissioned from the University of Nevada, Las Vegas International Gaming Institute. The report found that AI is "already embedded in core business functions", with advanced personalisation offering both benefits to customers and "potential increased risk of harm to vulnerable populations", and that many indicators recommended for detecting player risk lacked evidence, while payment-related indicators had the strongest. It recommended a task force, which the MGC formed and which Serpa now leads, according to Covers. Brodeur said Massachusetts is one of the only states, if not the only one, whose rules refer to AI at all, CDC Gaming reported.

Massachusetts licenses DraftKings, which is based in Boston, alongside the state's other online sportsbooks. In Washington, the SAFE Bet Act introduced by Senator Richard Blumenthal and Representative Paul Tonko would bar sportsbooks from using AI to track betting habits and personalise offers.

The Commission Has Chosen the Market-Wide Question Over the Single Case

By widening the inquiry to every licensee, the MGC has turned an allegation about one company into a question about industry practice, which is both fairer and more consequential. Fairer, because response modelling of the kind the Times described is common in consumer marketing, and there is no evidence yet that DraftKings is alone in using it. More consequential, because a finding that promotional AI is widespread would point to rules rather than to enforcement against one operator. The commission's own study supplies the tension it will have to resolve: it found the evidence for predictive harm indicators weak, which is DraftKings' stated reason for dropping its harm model, while finding that personalisation already carries risk. A regulator that accepts the first finding cannot easily ignore the second.

A Fact-Finding Exercise Is Not Yet Supervision

Maynard was careful to frame the step as information gathering, with policy to follow "when or if appropriate". That is the right order for a regulator facing an investigation it did not conduct, but it leaves the outcome open. The useful test is whether the staff review produces something operators must disclose, such as which variables drive promotional targeting and whether the same data is checked for signs of harm. Without that, the review will describe the practice without changing it. Massachusetts has the study, the task force and the rules to go further than most states, and it is the one regulator whose licensee is headquartered on its doorstep.

The commission has asked the right question of the whole market rather than one firm. What it does with the answers will show whether AI in gambling marketing is regulated or merely observed.