DraftKings has deployed a machine‑learning system that analyzes its own customers’ historical betting data to identify users who are statistically likely to lose money on wagers. The company trains the model exclusively on first‑party data—bet sizes, frequencies, outcomes, and promotional response logs—without purchasing third‑party datasets. Once the model flags a user as a probable loser, DraftKings serves that individual personalized promotions intended to encourage additional betting activity, effectively re‑engaging high‑risk gamblers. The AI component accelerates the processing of large behavioral datasets, allowing rapid iteration and refinement of targeting criteria, which in turn drives further data collection to improve model performance. Because the model operates as a black box, the specific features that influence its predictions are opaque, prompting continual accumulation of more granular betting data. This feedback loop amplifies the harms of behavioral advertising by sustaining a cycle where vulnerable players receive increasingly tailored incentives to keep gambling, while the harvested data may be shared with external entities such as insurers, banks, and law‑enforcement agencies. The approach illustrates how AI‑enhanced ad tech can exploit user vulnerability for profit without relying on external data brokers.