Online sports betting giant DraftKings is reportedly using artificial intelligence to identify and target chronic gamblers—specifically those likely to place losing bets—with personalized promotions designed to keep them engaged. The revelation, detailed by The New York Times and amplified by the Electronic Frontier Foundation (EFF), underscores the escalating harms of AI-driven behavioral advertising and has reignited calls to ban the practice outright.
Background & Context§
Behavioral advertising personalizes ads based on user data, and the more data a company collects, the more precisely it can target individuals. DraftKings, a major player in the online sports betting market, possesses vast amounts of betting records that reveal not only preferences but also vulnerabilities. According to the EFF, the company is leveraging this data to train machine learning models that predict which customers are most likely to place losing bets and respond to gambling promotions. This matters because it represents a particularly predatory application of AI: rather than mitigating harm, the technology is used to exploit it. As AI becomes more integrated into digital marketing, the ethical and regulatory stakes grow, especially in industries like gambling where addiction can have devastating consequences.
The News: What Happened Exactly§
According to a New York Times report, DraftKings is using its customers’ betting histories to train a machine learning model with a clear objective: find losing gamblers. Once identified, these customers are sent targeted advertising designed to lure them back to the platform to place more bets—bets that the model predicts will be losing ones. This is not incidental; it is a deliberate strategy. DraftKings has a direct financial incentive to keep losing gamblers active because they are the users who generate the most revenue. The company profits when these individuals bet and lose, making their continued engagement a business priority.
The EFF points out that this targeting disproportionately affects problem gamblers—people who repeatedly gamble despite harm to themselves, their finances, and their relationships. These individuals are highly likely to be flagged by the model because their betting patterns align with the characteristics of losing gamblers. By re-engaging them through promotions, DraftKings capitalizes on their vulnerability for profit instead of intervening to mitigate their risk. This is a classic case of online behavioral advertising taken to an extreme, where AI supercharges the harmful effects by automating and scaling the exploitation.
While predatory advertising isn’t new, the use of AI to process vast datasets and continuously refine targeting is a significant escalation. The EFF has long argued that all behavioral advertising should be banned, not just in gambling but across the digital economy. They contend that the fundamental issue is the collection and use of personal data to manipulate behavior, which inevitably leads to harm when applied to vulnerable populations. The DraftKings case provides a stark example: a company using sophisticated AI not to improve user experience or promote responsible gambling, but to maximise revenue from those least able to afford it.
The implications extend beyond DraftKings. If one company can deploy such a model, others in the gambling industry—and beyond—may follow. Regulators have been slow to address AI-driven behavioral advertising, often focusing on privacy rather than the algorithmic targeting of harm. The EFF’s call for a ban is a direct response to this gap, arguing that voluntary measures or transparency requirements are insufficient when the business model itself is predicated on exploitation. The report has sparked renewed debate about the ethical responsibilities of AI developers and the need for enforceable limits on how personal data can be used to influence behavior.
Historical Parallels & Similar Incidents§
The use of data to target vulnerable individuals is not unprecedented. In 2018, the Cambridge Analytica scandal revealed how Facebook data was harvested and used to psychologically profile voters and target them with political ads. That incident involved a third party exploiting a platform’s data policies, but the core mechanism was similar: using behavioral data to influence decisions, often without meaningful consent. Unlike DraftKings, Cambridge Analytica’s actions were not part of the platform’s intended business model, but the fallout led to significant regulatory scrutiny and changes in data protection laws. The lesson is that when data is used to manipulate, the harm can be widespread and systemic, and relying on self-regulation often fails.
Closer to the gambling industry, the case of UK betting firm William Hill offers a parallel. In 2020, the UK Gambling Commission fined William Hill £2.9 million for failing to protect consumers and for social responsibility failures, including not intervening when customers showed signs of problem gambling. While that case did not involve AI targeting, it highlighted how operators can prioritise profits over consumer protection. DraftKings’ use of AI to target losing gamblers is a technological leap beyond passive neglect—it is active pursuit. The contrast shows that as technology advances, the methods of exploitation become more precise and harder to detect, making regulatory oversight increasingly challenging.
These historical examples underscore a recurring theme: without strict regulation, companies will often push the boundaries of acceptable behavior to maximise profit. The EFF’s call to ban behavioral advertising echoes earlier calls to ban targeted political ads or to impose strict limits on data collection. The DraftKings case may serve as a catalyst for broader action, much like Cambridge Analytica did for privacy legislation. However, the outcome depends on whether policymakers recognise that AI-driven targeting is not just a privacy issue but a public health and safety issue, particularly when it comes to addiction.
# Hypothetical simplified model of a losing-gambler targeting system
import pandas as pd
from sklearn.ensemble import RandomForestClassifier
# Features: betting frequency, average bet size, loss rate, deposit patterns
data = pd.read_csv('user_betting_records.csv')
X = data[['bet_freq', 'avg_bet', 'loss_rate', 'deposit_volatility']]
y = data['is_losing_gambler'] # Label based on historical outcomes
model = RandomForestClassifier()
model.fit(X, y)
# Identify users with high probability of being losing gamblers
losing_gamblers = data[model.predict_proba(X)[:,1] > 0.8]
# Send targeted promotions to these users
for user in losing_gamblers['user_id']:
send_promotion(user, "Exclusive bonus for you!")This code snippet, while simplified, illustrates how straightforward it is to build a model that flags losing gamblers. The ethical line is crossed when such a model is used not to help but to exploit. The EFF’s stance is clear: the only way to prevent this harm is to ban behavioral advertising entirely, removing the incentive to collect and weaponise personal data. Whether regulators will heed that call remains to be seen, but the DraftKings case has undeniably brought the issue into sharp focus.