Loren H. | Tech & AI commentary contributor, Roar Tech Mental. Tested July 2026.
Governor Kathy Hochul didn’t mince words this spring. Her directive to the New York State Gaming Commission was explicit: prevent artificial intelligence from preying on gamblers. Not nudging them. Not optimising their experience. Preying. That’s a charged word, and she chose it deliberately.
For anyone who has spent time thinking critically about how AI recommendation systems actually work. Not how their product pages say they work. The framing wasn’t surprising. It was overdue. What is surprising is how little of the public conversation since has focused on the specific mechanism Hochul was warning about: the recommendation engine.
This piece isn’t about whether online casinos are good or bad. That debate is older than the internet. It’s about a narrower, thornier question. When an AI system tells a New York resident which casino to play at, what is it actually optimising for? And in a state where the legal framework for online casino gambling still doesn’t exist, who is checking?
What the Regulated Landscape Actually Looks Like
Before getting into the algorithm, it’s worth grounding this in what’s real. New York has legal sports betting. It does not have legal online casino gaming. Senator Joseph Addabbo’s SB 2614 has returned for another cycle in 2026, and the usual arguments are being rehearsed again in Albany. Nothing has passed.
That gap matters enormously. Players searching for online casinos in New York aren’t being directed to regulated, state-licensed platforms. Because those don’t exist yet. The most useful resource right now is an honest guide to New York online casinos, which maps what’s available versus what’s legal, what offshore platforms are accessible from the state, and what players should realistically expect given the current regulatory void. That kind of transparent, structured information is exactly what an AI recommendation engine is not built to provide.
AI recommenders don’t surface licensed-versus-unlicensed distinctions. They surface engagement
The Optimisation Problem Nobody Talks About
Here’s the thing most casino recommendation tools won’t tell you: the model is not trained to find you the safest casino. It’s trained to find you the casino you’re most likely to click on, sign up for, and return to.
Those aren’t the same objective. Not even close.
A peer-reviewed study published in Behavioral Sciences found that AI personalisation on gambling platforms directly shapes users’ risk perception. Not just their preferences. Players exposed to algorithmically tailored content reported higher confidence in their own gambling decisions, even when the outcomes didn’t justify that confidence. The system isn’t neutral. It’s building a self-reinforcing loop where the user feels more in control precisely because the interface has learned to make them feel that way.
That’s the baseline behavior of these tools in regulatedmarkets with licensed operators. In New York, where the online casino market is still offshore and largely ungoverned, those feedback loops have no floor.
What ‘Recommendation’ Really Means in an Ungoverned Market
Let’s be specific about what these tools are doing.
An AI casino recommendation engine ingests data signals. Search history, session duration, device type, time of day, geographic location, possibly social media activity depending on the permissions chain. And produces an ordered list of platforms. The ordering isn’t random. It reflects training data that was, in most cases, generated by affiliate marketing arrangements.
The casino that appears first isn’t necessarily the safest, the fairest, or the most compliant. It’s the one the model has learned is most likely to produce a conversion. In affiliate-heavy verticals, ‘conversion’ means a deposit.
None of this is secret. It’s how most recommendation systems in commercial contexts work. Amazon does it. Netflix does it. The difference is that a bad Netflix recommendation costs you two hours. A bad casino recommendation at 2 a.m. After a personalized bonus prompt has a different cost profile.
At SBC Summit Canada in May 2026, several industry speakers raised exactly this concern. That AI personalisation tools deployed in unregulated markets can be, in their words, “weaponised against vulnerable players.” The framing was stark enough that it made headlines. The follow-up action was less impressive.
The RNG Certification Parallel
Regular readers here will recognise the pattern. We’ve been through a version of this argument before with RNG certification and AI safety auditing. The question of whether the auditor has any real independence from the system being audited, and whether a certificate on a wall means what it claims to mean.
The same structural problem applies to casino AI recommenders. The entity generating the recommendation has a financial interest in the outcome of that recommendation. The oversight of that recommendation. The checks on whether it’s optimising for user welfare rather than affiliate revenue. Is either absent or self-reported.
For a deeper look at how that dynamic plays out in practice, the site’s earlier analysis of whether AI casino recommendation engines can actually be trusted is worth revisiting. The short answer, then and now, is: not without independent verification of what the model is optimising for.
Independent verification that, in New York’s current legal environment, nobody is requiring.
Why New York Is a Harder Case Than Most
Some states have created a workable framework. New Jersey’s Division of Gaming Enforcement requires licensed operators to disclose responsible gambling tools and submit to regular compliance audits. That’s not perfect, but it’s something. A licensed operator in New Jersey can’t just deploy whatever behavioral AI it wants without some regulatory surface area to push against.
New York has no equivalent for online casinos. The platforms accessible to New York residents from offshore jurisdictions are operating under Curaçao or Malta licensing. Frameworks that have their own enforcement gaps. And they’re under zero obligation to comply with anything Governor Hochul’s office says. Her directive applies to operators she can regulate. She can’t regulate platforms that don’t exist under New York law.
That’s the gap AI recommendation engines are filling. And they’re filling it profitably.
MIT Technology Review has spent considerable column space on the broader ethics of AI agents making consequential decisions autonomously. As contributor Will Douglas Heaven argued in late 2024, the core problem with AI agents isn’t capability. It’s accountability. Who is responsible when the agent makes a decision that harms someone? In the casino recommender context, that question has a clear answer: nobody, currently.
The Incentive Structure Is the Story
Step back from the technology for a moment. Strip the AI language out entirely.
You have a market. New York residents who want to play online casino games. That has no licensed domestic supply. You have offshore platforms competing for that demand. You have intermediaries (recommendation engines, affiliate sites, AI tools) who get paid when they successfully direct a user to one of those platforms. And you have zero independent oversight of how those intermediaries make their decisions.
Add AI back in. What changes? The intermediary is faster, more personalised, more persuasive, and less transparent about its decision-making. The incentive structure is identical. The harm potential is higher.
This is the argument Hochul is gesturing at, even if the directive focuses on licensed operators where she actually has jurisdiction. The real problem is the recommendation layer that operates before a player ever reaches a licensed or unlicensed platform.
What Would Actually Help
Clarity, mostly. Not optimism.
If SB 2614 passes and New York creates a regulated online casino market, the state will have the authority to set standards for how licensed operators use AI personalisation tools. That’s meaningful. It won’t cover offshore platforms that continue operating in the grey zone, but it would at least give regulators a foothold.
In the meantime, the most honest thing anyone navigating this space can do is use information sources that aren’t financially incentivised by the outcome of your casino choice. Structured editorial guides that separate licensed from unlicensed options, that explain what regulation actually covers and what it doesn’t, serve a different function than an AI tool trained on conversion data.
The EU’s AI Act simplification package, finalised by the Council in late June 2026, pushes high-risk AI systems toward greater transparency obligations. The United States has no equivalent at the federal level. New York, by passing real iGaming legislation, could create one at the state level. At least for the operators it can reach.
That’s not a sufficient solution. It’s a starting point.
—
Frequently Asked Questions
Is online casino gambling legal in New York in 2026? Online sports betting is legal in New York. Online casino gaming (slots, table games, live dealer) is not. Senator Addabbo’s SB 2614 has been reintroduced in 2026, but as of July 2026, no bill has passed. New York residents can access offshore platforms, but those operate outside state licensing.
What is Governor Hochul’s AI-and-gambling directive actually about? Hochul’s spring 2026 directive instructed the New York State Gaming Commission to develop rules preventing AI from being used to target vulnerable gamblers. It applies to licensed operators the state can regulate. Not to offshore platforms operating under foreign licensing frameworks that aren’t subject to New York law.
How do AI casino recommendation engines make money? Most operate on an affiliate model. When a user clicks through a recommendation and deposits on a platform, the recommender receives a commission. Typically a percentage of the player’s lifetime losses or a flat cost-per-acquisition fee. The model is trained, in effect, to maximise that conversion event, not to surface the safest or most player-friendly option.
Can AI personalisation actually change how people gamble? Yes. A 2025 peer-reviewed study in Behavioral Sciences found that algorithmically personalised content on gambling platforms altered users’ risk perception, making them feel more confident in their decisions regardless of actual outcomes. The effect was consistent across user groups and was stronger with higher personalisation intensity.
What should New York players look for when choosing an online casino right now? Look for platforms licensed under jurisdictions with meaningful enforcement (Malta Gaming Authority is generally stricter than Curaçao). Check whether the platform publishes its RTP data independently and has a visible responsible gambling mechanism. Don’t rely on an AI recommendation tool as your primary filter. Check what the tool is optimising for before trusting its output.
—
New York’s online casino debate will eventually resolve into legislation, one way or another. What won’t automatically resolve is the question of how AI recommendation systems operate in the space between demand and regulated supply. Hochul named the threat correctly. The harder work is building the accountability architecture that makes ‘preying’ measurable, attributable, and actionable. That work hasn’t started yet.
Gambling involves risk. Please play responsibly and only wager what you can afford to lose. If you feel gambling is becoming a problem, visit BeGambleAware.org or call 1-800-GAMBLER.
Loren Hursterer is the kind of writer who genuinely cannot publish something without checking it twice. Maybe three times. They came to expert analysis through years of hands-on work rather than theory, which means the things they writes about — Expert Analysis, Latest Technology Updates, Mental Health Innovations, among other areas — are things they has actually tested, questioned, and revised opinions on more than once.
That shows in the work. Loren's pieces tend to go a level deeper than most. Not in a way that becomes unreadable, but in a way that makes you realize you'd been missing something important. They has a habit of finding the detail that everybody else glosses over and making it the center of the story — which sounds simple, but takes a rare combination of curiosity and patience to pull off consistently. The writing never feels rushed. It feels like someone who sat with the subject long enough to actually understand it.
Outside of specific topics, what Loren cares about most is whether the reader walks away with something useful. Not impressed. Not entertained. Useful. That's a harder bar to clear than it sounds, and they clears it more often than not — which is why readers tend to remember Loren's articles long after they've forgotten the headline.
