New York Times reports DraftKings used AI to target promotions at losing gamblers
The New York Times reports DraftKings built an AI model to aim promotions at gamblers it expected to lose most, and shelved predictive work on problem gambling.
Why it matters
The claims touch advertising rules aimed at problem gamblers, a group state regulators have long targeted with restrictions on gambling promotions. If the internal models matched what the Times describes, DraftKings faces scrutiny from lawmakers and gambling regulators over how it uses player-level prediction data, and the episode may renew pressure on operators to disclose targeting practices. The desk has one source for the investigation, so the regulatory response is not yet visible in the record.
Where the sources disagree
Status of the problem-gambling prediction model
Used AI to aim promotions at gamblers likeliest to lose
Bitcoin.com News
Rejects any implication its marketing unfairly targets customers
Bitcoin.com News
The New York Times reports the responsible gaming risk-scoring project was shut down, while DraftKings has not confirmed that account in the material the desk holds.
Key facts
The New York Times reported that DraftKings built a machine-learning model that steered promotions toward gamblers it predicted would lose the most money.
Disputed figureReported by Bitcoin.com News
The New York Times reported that a 2023 model built for DraftKings' online casino division gave each player an elasticity score, where a higher score meant the player was likely to lose more money per promotion offered.
Reported by Bitcoin.com News
A data scientist began developing a predictive risk model for DraftKings' responsible gaming division in mid-2024, according to the New York Times report.
Reported by Bitcoin.com News
The New York Times reported that the problem-gambling prediction project was shut down, and former staff said such models were shut down or shelved.
Disputed figureReported by Bitcoin.com News
DraftKings' responsible gaming division had about 50 people within a workforce of 5,000, the New York Times reported.
50Reported by Bitcoin.com News
The Times based its investigation on interviews with more than 40 former DraftKings employees, internal research memos, presentations, Slack messages and betting records from company experiments on customers.
40Reported by Bitcoin.com News
DraftKings rejects the New York Times characterisation and rejects any implication that its marketing unfairly targets customers.
Reported by Bitcoin.com News
The Times described the risk model as intended to predict harmful behaviour days or weeks ahead, which the data scientist working on it was developing for the responsible gaming division.
Reported by Bitcoin.com News
What happened
The New York Times reported on Saturday that DraftKings built a machine-learning model that scored each online casino player on how much money they were likely to lose per promotion, and steered marketing toward the highest-scoring players. The same investigation says a data scientist began building a predictive model for the responsible gaming division in mid-2024 to flag customers at risk of harmful behaviour days or weeks ahead, and that project was shut down. The Times based its account on interviews with more than 40 former Draftkings employees plus internal research memos, presentations, Slack messages and betting records from experiments run on customers. The Times described two AI-assisted buildouts that developed in opposite directions. Draftkings rejects the characterisation and disputes any implication that its marketing unfairly targets customers. No other outlet in the desk's file has covered the investigation, so the account rests on the Times reporting and Draftkings' response to it.
How the story developed
- First report
First report by Bitcoin.com News
Draftkings Shelved a Problem-Gambling AI While Targeting Likely Losers
- Update
DraftKings builds elasticity model for online casino division
A 2023 model gave each player an elasticity score indicating how much money that player was likely to lose per promotion.
- Official statement
DraftKings rejects the characterisation
The company rejects the Times account and any implication that its marketing unfairly targets customers.
- Update
Responsible gaming risk model project begins
A data scientist started developing a model to score customers for risk of harmful behaviour, for a division of about 50 people inside a 5,000-strong workforce.
- Update
New York Times publishes investigation
The Times reported the two AI buildouts and their opposite directions, drawing on more than 40 former employees and internal documents.
Affected entities
Original sources
| Publisher | Report | Role | Published |
|---|---|---|---|
| Bitcoin.com News Company, official | Draftkings Shelved a Problem-Gambling AI While Targeting Likely Losers | Original |