New York advances AI betting rules as data-sharing gap widens

New York's AI betting rules advance as problem gambling data-sharing gap widens

New York State Gaming Commission Chairman Brian O'Dwyer told the nine licensed sportsbook operators in the room something they already knew was coming, and then made it sharper. If the industry does not solve the problem of a restricted bettor moving from one platform to another, he said, the commission would solve it for them — and they might not like the solution.

That is not a threat buried in regulatory language. It is the chairman of a licensing body telling operators that their licences are contingent on behaviour the licences do not yet require. The gap between those two things is where this story sits.

The commission voted unanimously to advance two proposals. The first restricts how mobile sports betting operators use artificial intelligence to personalise promotions, odds, betting limits, and product recommendations. The second requires licensees to identify at-risk bettors and designate a person to engage them directly. Both proposals go back for public comment before a final vote.

The AI restriction is modelled on New Jersey's framework, which does not prohibit the technology outright but bars its use for targeted personalisation of the kind that a problem gambler is most vulnerable to — a promotion calibrated to a losing streak, a limit adjusted to keep a high-volume bettor in the game. The specific systems named in the proposal include automated-decision systems, algorithmic profiling, machine learning, and generative AI. That list is broad enough to reach most of what the major platforms currently run.

The data-sharing question is the harder problem. O'Dwyer's concern is structurally correct: a bettor flagged by DraftKings can open an account with FanDuel before the ink dries. Nine separate licence holders, nine separate risk databases, no obligation to share. The proposal requires each licensee to designate a responsible gambling contact, but it stops short of mandating cross-operator data sharing — which is precisely the mechanism that would close the gap O'Dwyer described.

I have watched this pattern in other regulated markets. The regulator identifies the structural failure accurately, advances a rule that addresses part of it, and then issues the warning that the rest must come from industry coordination. Industry coordination on data sharing, in a competitive market, moves slowly or not at all. The warning is genuine. The timeline for acting on it is not yet set.

The commission's AI rule follows Governor Kathy Hochul's March programme on underage gambling, at-risk marketing, and addiction. That origin matters because it signals the political durability of these proposals — they are not a commissioner's initiative, they have executive backing, and the nine operators know it.

Kalshi and Polymarket are not touched by this directly. These are sportsbook rules, applied to state-licensed mobile operators. But the AI personalisation question will reach prediction markets eventually, and New York's framework — once finalised — will be the template other states examine. The commission's decision to advance both proposals unanimously, without a dissenting voice among its members, is the data point that matters most here.