Bettor Sues DraftKings, Alleging Its AI Model Steered Promotions to Likely Losers

A proposed class action filed in Boston accuses DraftKings of using customers' data to find the gamblers most likely to lose, then sending them more offers, in breach of its own privacy promises. It leans heavily on the New York Times investigation, and DraftKings says it does not use AI that way.

• Daniel Vest, a West Virginia resident, filed a proposed class action against DraftKings in the US District Court for the District of Massachusetts on 30 September, case no. 1:26-cv-14462, assigned to Judge Brian E. Murphy
• The complaint alleges, largely on the strength of a New York Times investigation, that DraftKings built a machine-learning model to identify customers likely to lose more in response to promotions and targeted them with extra offers and messages
• It claims breach of contract, breach of implied contract and money had and received, arguing that DraftKings' privacy notices promise to use customer data to check that people are "responsibly playing"
• Vest says he received at least about 70 emails, texts and notifications from DraftKings in the 30 days to 25 September; he seeks damages, refunds, disgorgement and an injunction against the use of the model
• DraftKings says it "does not use AI to target customers based on losses", and the Massachusetts Gaming Commission is already reviewing how licensed sportsbooks use AI

A Privacy Notice Becomes the Contract in Dispute

The complaint was filed on Wednesday 30 September by Jason Leviton and Brendan Jarboe of Block & Leviton LLP, on behalf of Vest "and all others similarly situated". It invokes federal jurisdiction under the Class Action Fairness Act, alleging that more than $5 million is in dispute, and demands a jury trial. DraftKings is headquartered in Boston.

Its central allegation is that DraftKings "has weaponized AI" to "understand and exploit users' vulnerabilities". Most of the factual detail comes from the New York Times report of 19 September, which the complaint quotes repeatedly: that in 2023 DraftKings built a machine-learning model to work out which customers were more likely to respond to promotions by gambling, and losing, more, that it scored customers so that a higher score meant more expected loss per promotion, that users with below-average scores were deemed "inelastic" and marked for fewer incentives, and that a separate model to flag customers at risk of harm was sidelined. On that basis, and "upon information and belief", the complaint alleges that customers deemed "elastic" were sent extra emails, texts, in-app notifications and promotions intended to keep them gambling, without being told their data was used that way.

The legal theory rests on DraftKings' own documents. Its sportsbook and casino privacy notices, the complaint says, state that "as part of our commitment to responsible play, we may use the information collected about you to assess whether you are responsibly playing" and may contact customers who show "potentially problematic information". Vest argues that customers accepted that promise by using the service and paying DraftKings, and that the company broke it by using the same data "for the exact opposite purpose".

Seventy Messages in a Month, and a Massachusetts Rule

Vest, according to the complaint, has used DraftKings' sportsbook and online casino, has gambled "thousands of dollars" with it each year for several years, and received at least about 70 emails, texts, notifications and messages in roughly 30 days before 25 September, many of them offering a promotion. The complaint reproduces several of them. It does not allege that Vest himself was scored by the model; it proposes a class of everyone DraftKings deemed "elastic" or who received targeted messages "based on DraftKings' use of the Model", a group it says is likely to number at least in the thousands and could be identified from DraftKings' own records.

The complaint also cites Massachusetts' sports wagering data privacy rule, 205 CMR 257.02, first adopted in 2023, which bars an operator from using a patron's personal or confidential information to promote wagers or offers based on any "machine learning, artificial intelligence, or similar system that is known or reasonably expected" to make the platform "more addictive". For now Vest pleads only common-law claims. The complaint says he has sent DraftKings a demand letter under the state's consumer protection law, Chapter 93A, and will seek to add that claim after 30 days unless the company makes an acceptable settlement offer. The relief sought includes damages, the return of money acquired through the model, disgorgement, statutory damages and an injunction barring DraftKings from using the model to target users.

DraftKings rejected the premise. "DraftKings does not use AI to target customers based on losses, nor do we use AI to target customers based on indicators of potential problem gaming," spokesperson Park Winslow said, according to WBUR, adding that the company would vigorously defend any lawsuits. The company told the Times last month that promotions go to customers with "sustained, engaged use" and that the 2023 test was "preliminary and inconclusive". Massachusetts Attorney General Andrea Campbell's office said the allegations "raise serious concerns", and State Auditor Diana DiZoglio called them "deeply concerning and, if substantiated, completely unacceptable", WBUR reported.

The Complaint Is a Bet on Discovery

Almost every factual allegation about the model is drawn from the Times, which relied on former employees and internal documents; the complaint cites no other source for how the model worked. That is a weakness at the pleading stage: DraftKings can argue that a complaint built on a newspaper's account, and pleaded on information and belief, does not show that this plaintiff was scored or harmed by the model. It is also the point of filing: if the case survives a motion to dismiss, discovery would reach the model, its scores and the promotions sent on them. The class definition depends entirely on records only DraftKings holds.

A Privacy Notice Was Not Written to Be a Promise

The breach-of-contract theory is novel and exposed. The notice language Vest relies on is permissive: DraftKings "may" use data to assess responsible play and "may" contact customers. Operators will argue that such wording describes permitted uses rather than undertakes a duty, and that a privacy notice is not a bargained-for contract. But the theory turns standard responsible-gambling language into a measurable commitment, and if a court accepts that it has content, every operator whose privacy notice makes the same promise while its marketing team models customer value faces the same question.

Massachusetts Has the Only Rule That Fits, and It May Not Reach This Plaintiff

The Massachusetts rule on addictive AI is unusually specific, which is why the complaint cites it, but it governs sports wagering in Massachusetts. Vest lives in West Virginia, and the model the Times described was first built for online casino, a product Massachusetts does not license, although the paper reported similar models for sports betting. Its relevance is less as a cause of action than as evidence of a standard: a regulator in DraftKings' home state wrote down in 2023 that AI known to make a platform more addictive may not drive promotions. The commission's review, and any Massachusetts enforcement that follows, may matter more than this suit.

The suit adds a contract claim to the addiction suits operators already face. Its fate depends less on what the Times reported than on what DraftKings' own data show.