A Federal Licence Did Not Stop New York

Two lawsuits, filed in opposite directions

On Friday 31 July 2026, Governor Kathy Hochul and Attorney General Letitia James announced that New York had sued Kalshi for running an illegal gambling operation. The state alleges the company accepts wagers without registering with the New York State Gaming Commission, that the outcomes traded on its platform turn more on chance than on skill, and that it permitted users aged 18 to 20 to trade in a state that requires participants to be at least 21. James put the state's position in one sentence: "No matter what they call themselves, prediction markets like Kalshi are gambling platforms, plain and simple."

On the same day, the federal government went to court against New York. The Commodity Futures Trading Commission and the United States asked a separate federal court to temporarily prohibit New York from bringing or continuing enforcement actions against Kalshi and other CFTC-regulated companies. Their argument is that the Commodity Exchange Act gives the CFTC exclusive jurisdiction over event contracts listed on federally regulated exchanges, and that this preempts state gambling law.

This is not the first round. On 8 July 2026 a federal judge rejected Kalshi's own attempt to block New York's gambling enforcement, and the company appealed. What changed on 31 July is who is carrying the argument. Kalshi lost the motion when it argued preemption itself; three weeks later its regulator is arguing it instead, and against the state directly.

The number is built from the whole business

New York is asking for a court order halting the operation, forfeiture of all gains from it, fines equal to three times those gains, restitution to users who traded on the platform, and a penalty of $100,000 for each attempt to offer sports wagering. Those components produce the figure being reported around the case: no less than $36 billion. Restitution is sought for users nationwide, not only for users in New York.

That is the mechanical difference between a compliance failure and a classification failure. A compliance failure is priced against the breach: a late filing, a missing control, a fine proportionate to the lapse. A classification failure is priced against the activity, because if the conduct was never lawful then every unit of revenue it produced is treated as a gain to be given back. The multiplier does the rest. Three times an illegal gain is a different order of exposure from a percentage of turnover.

Hochul framed it as a level-playing-field point, saying Kalshi "has chosen to ignore New York's gaming laws, which exist to protect consumers, prevent problematic gambling, deliver funding for critical public services, and ensure that every company plays by the same rules." Whatever a court eventually decides, the structure of the claim is the lesson. The size of the demand is a function of how long the product operated, not of how badly it behaved.

Europe answered the classification question in July

The European Securities and Markets Authority published a public statement on 3 July 2026 on whether national product intervention measures covering binary options apply to event contracts. Its answer was that not every event contract is a financial instrument, but that those whose event question relates to an underlying named in Section C(4) to (10) of Annex I of MiFID II do qualify, and that they are derivatives because the payout is binary. A fixed sum or nothing, decided by a yes-or-no outcome, is the shape the rules already describe.

Once a contract lands inside that definition, the retail market is closed by default. The national measures that followed the European ban on binary options for retail clients prohibit marketing, distributing or selling those products to retail investors. Offering the qualifying contracts to professional or institutional clients remains possible, but doing so requires authorisation as an investment firm under MiFID II. Neither route is available to a platform that holds no European authorisation at all.

The Commission is separately consulting on how prediction contracts should be treated as part of its review of the markets in crypto-assets regime, with responses open until 30 September 2026. So the European position is not finished. It is, however, already operative, which is the part that matters for anyone building on the assumption that a rulebook is only real once someone has been sued under it.

One product, two regulators, and a licence that answers neither

Nine European gambling regulators, from Belgium, France, Germany, Italy, the Netherlands, Poland, Portugal, Spain and Switzerland, announced coordinated action against unlicensed prediction market platforms on 17 June 2026, timed to the start of the football World Cup. They cited round-the-clock access, absent stake and time limits, weak age and identity checks and the risk of harm among young adults. Several national authorities had already acted: Polymarket has been blocked in France, Belgium, Poland, Portugal, Romania and Hungary, and the Spanish regulator halted both Polymarket and Kalshi for operating without a licence.

Put the securities answer and the gambling answer side by side and the structural problem appears. A European authorisation as an investment firm would satisfy the securities regulator and would not answer the gambling regulator, which is applying a different statute to the same transaction. A national gambling licence would work in the other direction. The two classifications are made independently, by different authorities, against different definitions, and neither is required to defer to the other. As of the middle of this year, none of the major platforms operated under a licensed European entity.

The American fight is about which regulator owns the product. The European position is harder than that, because it does not require anyone to win: the same contract can be a financial instrument to one authority and gambling to another at the same time, and an operator can be exposed on both counts while holding a licence that is genuine. Kalshi is licensed. That is not in dispute. What the New York case tests is whether being licensed by one authority is an answer to a second authority that never issued it.

What to check before your product straddles a definition

This reaches well past prediction markets. The same structure appears wherever a product sits between two rulebooks: a yield feature that a securities regulator may read as a fund, a guarantee that an insurance supervisor may read as underwriting, a deferred payment that a credit authority may read as lending, a workflow tool that becomes a medical device the moment it suggests a diagnosis. In each case the operator holds a real licence and the question is whether it covers the conduct a second regulator is looking at.

The practical test is not whether you are licensed but how many authorities can reach the same conduct. Write the list. For each one, note the definition it would apply, whether your existing authorisation speaks to that definition, and what the remedy looks like if the answer is no. Where the remedy is disgorgement of gains rather than a fine, the exposure grows with every month of ordinary trading, which means delay is not neutral.

The second question is jurisdictional reach. New York is seeking restitution for users nationwide from a business it says was never lawfully offered in the state. Under European rules the equivalent trigger is whether you are marketing into a member state at all, not where you are established. If your growth plan involves accepting customers from a country where you hold nothing, the classification question there is already live, whether or not anyone has written to you about it.

Read next: Your Vendor Decides When Your 72 Hours Start  |  Today's Repair Duty Covers Devices You Already Sold