DraftKings is training machine learning models on its customers' betting histories to identify the users most likely to lose — and then targeting those users with promotions engineered to bring them back to the platform. The New York Times reported the practice, and the Electronic Frontier Foundation is using it as a case study for why all online behavioral advertising should be banned. The mechanism is straightforward and ugly. DraftKings collects first-party data on every bet its users place. The ML model identifies patterns associated with losing gamblers — the very users who generate the company's profit margin. Those users then receive personalized promotions designed to re-engage them. The population most likely to be flagged by this model overlaps heavily with problem gamblers: people who continue betting despite financial, personal, and psychological harm. This is not a novel category of harm. Online behavioral advertising has always worked by collecting data about user behavior and using it to serve ads calibrated to exploit that behavior. What AI changes is scale, speed, and opacity. Models process vastly more data points than human analysts, find patterns humans would miss, and operate as black boxes where even the engineers building them cannot reliably predict which data features the model weights most heavily. The incentive to collect more data intensifies because the model's appetite for training data is effectively unbounded. The EFF makes a structural point about first-party data that policymakers need to hear. DraftKings appears to be using only data it collects directly from users — no third-party data purchases required. This means regulatory frameworks focused exclusively on restricting third-party data sales would not touch this practice. A company with a large enough user base and enough behavioral signal from its own platform can build highly effective predatory targeting systems entirely within the walls of its own data collection. The downstream surveillance implications extend beyond gambling. Data collected for ad targeting is routinely sold to insurance companies, banks, and law enforcement agencies including CBP. ICE published a Request for Information earlier this year explicitly seeking to understand how commercial ad-tech data could support its investigations. The data pipeline built for personalized advertising doubles as infrastructure for state surveillance. The EFF's proposed solution is a ban on behavioral advertising entirely. The logic: if companies cannot serve personalized ads, the economic incentive to collect the behavioral data powering those ads collapses. Contextual advertising — ads placed based on the content of the page rather than the profile of the user — would remain viable. The EFF also points users to its Surveillance Self Defense project and mobile app privacy guides as interim measures. The DraftKings case is clarifying because the extraction is so legible. A company identifies its most vulnerable customers using their own behavioral data, then uses AI to ensure those customers keep coming back. The question for regulators is whether the answer is disclosure requirements, algorithmic audits, or the structural remedy the EFF advocates: removing the economic incentive entirely by banning the practice.
