By Stuart Kerr, Technology Correspondent, LiveAIWire
AI in sports betting has become a genuine double-edged sword, and the industry’s own researchers are the ones saying so plainly. The same multimodal AI systems that let sportsbooks detect a customer sliding toward problem gambling, days or weeks before a human reviewer would notice, are built on platforms explicitly optimised to maximise engagement. “We know that the application of AI in this field also is a double-edged sword,” Dr Zahra Shakeri, an assistant public health professor at the University of Toronto, said in comments reported by the Washington Examiner. “There should definitely be policy regulations to target both how AI is being used and how AI can be used to address the problem.”
That tension, detection tool and engagement engine built from the same underlying technology, is now the central regulatory question in an industry moving faster than the rules meant to govern it.
How AI in Sports Betting Actually Spots a Problem Before Humans Do
Online gambling platforms generate a continuous stream of behavioural data, deposits, cancelled withdrawals, session length, customer support messages, and multimodal AI systems can now analyse all of it together rather than reacting to a single flagged transaction. Shakeri described the shift plainly: traditional tools see the bets, multimodal AI sees the story around the bets, which is where harm shows up early. “If we only look at transactions using traditional techniques, we catch the harm too late,” she said. “It’s like diagnosing lung disease only by checking how often someone buys cough drops.”
Fanatics Sportsbook has built exactly that kind of system into its responsible-gaming programme. The company uses Neccton, an AI-based risk-scoring tool developed by gambling researcher Dr Michael Auer, which evaluates dozens of behavioural indicators and assigns each customer a risk score. Anthony D’Angelo, Fanatics’ head of responsible gaming, told the Washington Examiner that a pattern such as frequently cancelled withdrawals or repeated insufficient-funds attempts drives a customer’s score up, prompting a human review rather than an automatic account action. “The AI does not diagnose addiction or act autonomously,” D’Angelo said. “It standardises reviews and reduces subjectivity.”
The Same Technology Can Be Used to Target Vulnerable Bettors
The risk researchers keep returning to is that a model trained to identify behavioural vulnerability can, in principle, be repurposed to identify exactly which customers are most responsive to incentives or promotions. Shakeri, who has worked on teams building multimodal gambling-risk systems, said that overlap is not hypothetical. Dr Kasra Ghaharian, director of research at UNLV’s International Gaming Institute, added that the industry’s next phase will likely bring more AI-driven betting assistants directly into consumer hands, blurring the line between an AI that guides a bet and one that influences it. “I don’t think gambling operators might even be aware that these technologies are being used and what the risks are,” Ghaharian said.
Regulators Are Moving to Ban the Targeting Use Case Outright
New York has become the first state to propose a direct answer to that overlap. As part of a 2026 State of the State initiative, Governor Kathy Hochul directed the New York State Gaming Commission to draft rules that would bar every licensed gaming operator in the state, not sports wagering alone, from using AI to offer personalised promotions or suggested wager amounts to a customer. The draft regulations pair that ban with thirteen specific behavioural triggers requiring operator intervention, including deposits exceeding 10,000 dollars in 24 hours, three cancelled withdrawal requests in ten days, and a customer’s account time increasing by 50 percent or more week over week.
Each trigger escalates through a defined three-phase response: an initial outreach pointing the customer to responsible-gaming tools, a mandatory harm-awareness video if the behaviour continues, and a direct call from a designated responsible gaming lead who can suspend the account or refer the customer to treatment. New York’s proposal treats the AI-targeting ban and the intervention framework as two sides of one policy, permitting AI to flag risk while explicitly prohibiting the same technology from being used to keep a flagged customer engaged.
Why Legislators in Other States Are Watching Closely
New York is not acting alone, though it has moved furthest. Massachusetts Gaming Commission chair Jordan Maynard has argued publicly that technology capable of targeting bettors could just as easily be redirected to promote healthy behaviour, and that regulators need to get involved precisely because operators will not make that trade-off voluntarily. Illinois has considered legislation that would bar betting apps from collecting data with the specific intent of predicting how a customer will gamble in a given scenario, and separate proposals in the state would restrict the push notifications and reminders that AI-driven personalisation engines use to bring a lapsed bettor back to the app.
The pattern across these proposals is consistent: lawmakers are not trying to ban AI from sportsbooks. They are trying to separate the detection use case, which most stakeholders agree has genuine safety value, from the personalisation use case, which the same underlying models make more precise and more persuasive at exactly the moments a vulnerable customer is least equipped to resist it.
Why This Fits a Pattern Beyond Gambling
The core problem sportsbooks are wrestling with, a system built to detect risk sitting on the same infrastructure as a system built to maximise engagement, is not unique to gambling. LiveAIWire’s coverage of the AI credit score’s persistent lending gap found a related dynamic in consumer finance: a scoring model built to serve a legitimate risk-assessment purpose carries the same underlying behavioural data that could, with different incentives applied to it, be used against the consumer it was built to evaluate fairly. The industries differ, but the structural question, whether a behavioural model is serving the person it is watching or the platform that built it, recurs whenever AI sits this close to a product engineered for engagement.
LiveAIWire’s broader reporting on how much AI should know about citizens found the same accountability gap playing out at a larger scale: systems capable of continuous behavioural monitoring tend to expand into new uses faster than the rules meant to constrain those uses, precisely because the technical capability arrives well before the regulatory framework does. Sports betting compresses that same gap into a market where the underlying product is designed to be habit-forming by commercial necessity, which is exactly why New York’s regulators are trying to draw a bright line now rather than after the technology has matured further.
What This Means for Anyone Using These Platforms
For bettors, the practical takeaway is that the responsible-gaming tools now built into most major sportsbooks, deposit limits, cool-off periods, self-exclusion, are running on the same AI infrastructure as the personalisation engine encouraging continued play, and the two systems are not always working in the bettor’s interest to the same degree.
Anyone who notices their own betting patterns matching the kind of triggers regulators are now writing into law, escalating deposits, chasing losses, cancelling withdrawals, has a concrete reason to use those tools proactively rather than waiting for an operator’s intervention system to flag it first. The same caution LiveAIWire has urged around AI systems built to read and respond to human behaviour at scale applies here too: a system this good at reading behavioural signals is exactly as good whether the purpose behind it is protective or extractive, and the purpose is set by the operator, not the technology. New York residents can reach the state’s confidential problem gambling helpline, HOPEline, at 1-877-8-HOPENY.
For the industry, the direction of travel looks increasingly like New York’s model: permit AI where it demonstrably protects customers, and draw an explicit, enforceable line around AI’s use where it demonstrably serves the platform’s engagement metrics instead. Operators that build that separation into their systems before it is mandated are likely to face a far smoother transition than those waiting for a regulator to force the distinction on them.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.