New York regulator threatens action over sportsbook data failures

Brian O'Dwyer did not leave much room for interpretation. When the New York State Gaming Commission chair told the nine licensed sportsbooks that they were "on notice," he followed it immediately with the detail that mattered: if the industry does not solve the problem of at-risk bettors moving between operators, the commission will solve it for them, on terms the operators may not find comfortable.

That is not a regulatory consultation. That is a deadline without a date.

The proximate issue is narrow but genuinely hard. A customer flagged by DraftKings as a problem bettor can close the app and open FanDuel. The commission's proposed rule would require each licensee to designate someone responsible for intervening with at-risk patrons, but O'Dwyer made clear that individual operator compliance misses the point if the data does not move across the industry. The nine operators — Bally Bet, BetMGM, Caesars, DraftKings, Fanatics, FanDuel, Resorts World Bet, Rush Street Interactive, and theScore Bet — have maintained that responsible gaming is already integral to how they operate. O'Dwyer's remarks suggest the commission has looked at the evidence and is not persuaded.

The AI piece is the more commercially consequential proposal. The commission advanced rules that would bar sportsbooks, their affiliates, and vendors from using machine learning, algorithmic profiling, or generative systems to personalize promotions, odds, betting limits, or product recommendations at the individual level. New York modelled this on New Jersey's approach, which restricts rather than prohibits, but the direction of travel is unmistakeable.

This is where I think consensus is reading too narrowly. Most of the commentary has treated New York's AI rules as a consumer protection measure with limited commercial bite — the kind of regulation that generates compliance costs without fundamentally altering the business. I disagree. The personalization engine is not a feature of the sportsbook product. It is increasingly the product. The operators who have spent the last four years building customer lifetime value models around individualized offers are now being told that the most profitable application of those models — targeting the bettors most likely to respond — is precisely what the regulation will prohibit. If the rule advances to a final vote in its current form, it does not inconvenience the business model. It cuts through the center of it.

What matters for market positioning is the sequence. The revised AI proposal goes back for additional public comment before any final vote, which means the operators have time to negotiate the edges. O'Dwyer's data-sharing ultimatum has no stated timeline, which is typically how regulators move when they want the industry to act before the rule is written. The threat of an imposed solution is more useful to a regulator than the solution itself.

I have watched this pattern before — a regulator who cannot yet compel an outcome stating loudly that he would prefer not to compel it. The industry reads it as a negotiation. Sometimes it is. O'Dwyer's specific language, "they may not like or the solution we come up with," does not read like someone who expects to be talked out of it.

The nine operators now have a collective action problem. Agreement on a shared exclusion database is commercially rational for the industry as a whole and individually costly for whoever has the most to lose from transparent data-sharing. That is not a problem the market resolves quickly on its own, which is exactly why O'Dwyer said what he said.