Online platforms have spent years perfecting user engagement through behavioral advertising, leaning on mountains of telemetry to personalize promotions. But as an analysis published by Devanshi Nishar demonstrates, the mechanics have evolved from simple ad-serving into algorithmic exploitation. According to reporting from the New York Times, online sports betting operator DraftKings deploys machine learning models directly against its user base, training them on internal betting records specifically to identify and isolate users who consistently lose.
First-Party Optimization of Losing Patterns
Once these patterns are flagged, DraftKings deploys targeted advertisements engineered precisely to pull these individuals back in for more wagers.
DraftKings is using AI to target customers who are most likely to place losing bets and respond to gambling promotions.
Because these vulnerable users generate the bulk of platform revenue, the algorithm automates re-engagement exactly where players exhibit persistent losing behavior. Instead of mitigating risk or adding friction, the system capitalizes on high-risk profiles, turning predictive modeling into a weaponized retention tool.
The Expansion of Behavioral Tracking Infrastructure
This is where the illusion of standard data protection shatters. Traditional privacy compliance—treating user data as a static asset to be anonymized or scrubbed—offers zero defense when the liability stems from how inference engines actually behave. Ad tech systems create vast data footprints that stretch far beyond simple consumer interfaces, and regulators are shifting rapidly toward aggressive algorithmic audits that treat predictive targeting as an active regulatory violation.
For engineering teams and technical leads evaluating internal recommendation engines, the mandate this week is straightforward: review whether your active retention models optimize against user harm metrics, or if you are simply building a compliance time bomb disguised as personalization.