DraftKings Trained AI to Find Losing Gamblers, Then Targeted Them
EFF says the company runs machine learning on customers' own betting records to flag the ones most likely to keep losing money, then sends them promotional ads. The fix lawmakers keep proposing, restricting third-party data, would not stop any of it.
Explore Strange New Parts ↓
DraftKings trained AI on its customers' own betting histories to identify the ones losing money ... and then it sent them ads to bring them back.
That is the finding at the center of a new Electronic Frontier Foundation analysis published September 24. The company trains machine learning models on customers' own betting records to identify the ones most statistically likely to keep losing. It then sends those customers promotional advertising designed to pull them back to the platform.
Problem gamblers are, by definition, people who repeatedly gamble despite negative financial and personal consequences. They are also, by EFF's analysis, the precise population a model trained to find losing bettors would surface first.
The Loop the Algorithm Builds
The model does not consult third-party data brokers or purchased audience profiles. It draws entirely on first-party behavioral data, the betting histories DraftKings collects directly from its own users. That distinction matters enormously for the policy debate happening in state legislatures and at the FTC.
Most proposed reforms target third-party data sales. DraftKings' system, as EFF describes it, would not be touched by any of them. The company built a vulnerability detector using only the information its customers handed over by placing bets.
"Online behavioral advertising incentivizes the collection of vast quantities of data to power ad tech," EFF wrote. "Adding AI into the mix means that even more data is collected to train and refine models."
The Data Does Not Stay at DraftKings
There is a second layer EFF flags. The behavioral data powering ad targeting does not live exclusively in the advertising ecosystem. CalMatters has reported that ad tech data is sold to insurance companies, banks, and financial services firms.
The government is also a buyer. ICE published a Request for Information seeking ways that "commercial Big Data and Ad Tech providers can directly support investigations activities." A bettor's profile built for ad targeting can become a law enforcement data point.
At a glance
- The model
- DraftKings trains machine learning on customers' betting records to identify losing gamblers and targets them with promotional ads, per EFF.
- Data type
- First-party only. No third-party data brokers involved, which means most proposed regulatory fixes would not apply.
- Vulnerable target
- Problem gamblers (people who repeatedly gamble despite harm) are the population most likely to be flagged by a model trained on losing behavior.
- Government pipeline
- ICE has sought RFI on whether commercial ad tech data can support investigations. Ad tech data also flows to insurers and banks, per CalMatters.
- EFF position
- Ban all behavioral advertising to eliminate the financial incentive to collect behavioral data, regardless of whether it is first- or third-party.
EFF's Ask Is Bigger Than a DraftKings Fix
The organization's position is not a call for DraftKings to change one algorithm. EFF argues that all behavioral advertising should be banned outright. The logic: so long as ad revenue depends on behavioral targeting, every company in the space has a financial incentive to collect maximum data, whether from third parties, from first parties, or from AI models trained on what users do inside a platform.
Removing the profit motive is the structural fix. Nothing else closes the loop.
DraftKings did not respond to EFF's request for comment at the time of that publication. Ad Watch News also reached out to DraftKings; no response was received by publication time.
The FTC has active rulemaking on commercial surveillance and data security. Its next public comment window had not been posted as of this writing.
Ad Watch News is an independent commentary site with no affiliation to DraftKings, the Electronic Frontier Foundation, or any other organization named here. This analysis draws from EFF's reporting, linked below.
