On 7 July 2026 a federal judge in Manhattan told Kalshi that New York may enforce its gambling laws against sports event contracts. Twenty days later a federal judge in Minnesota told the same company, in consolidated cases brought alongside Polymarket US and the CFTC itself, that a state ban on the same category was likely preempted by federal law. Neither ruling was sloppy. Both courts read the Commodity Exchange Act closely and came out in opposite places on what it does to state gaming statutes.
For anyone running a prediction market, that is not a headline. It is an operating condition, and by mid August 2026 it had lasted long enough that "we are waiting for legal clarity" stopped being a plan and became a way of postponing one. This piece is about the question that does not get written up: what a serious operator does with a map that contradicts itself, and why the answer is less dramatic and more expensive than it sounds.
Three weeks, two federal courts, opposite answers
In KalshiEX LLC v. Williams (No. 1:25-cv-08846, S.D.N.Y.), Judge Analisa Torres denied Kalshi's motion for a preliminary injunction against the New York State Gaming Commission on 7 July 2026. In the reading of Shook, Hardy and Bacon's analysis published on 24 July 2026, the court rejected field preemption because the CEA's savings clause and the Special Rule's own reference to state law violations suggest Congress meant the two systems to coexist, and rejected conflict preemption because Kalshi could in principle hold a New York licence and still meet the CFTC's impartial access requirement. Gambling regulation, the court held, sits squarely inside traditional state police power, which carries a presumption against preemption. The same firm reports the opinion treating self certification as reflecting only the exchange's own view of legality rather than agency approval, and CFTC inaction under the Special Rule as no proof that a contract is lawful.
Then on 27 July 2026, in consolidated proceedings covering United States and CFTC v. State of Minnesota, KalshiEX LLC v. Ellison and QCX LLC v. Ellison (Nos. 26-cv-2661, 26-cv-2778 and 26-cv-2841, D. Minn.), Judge Katherine Menendez granted a preliminary injunction days before Minnesota's new criminal statute was to take effect. Her 44 page order, published in full on govinfo, rested on four things: the CEA's grant of exclusive jurisdiction over transactions involving swaps traded on designated contract markets, the breadth of the statutory definition of a swap, the interplay between the Special Rule and the categories Minnesota tried to ban, and recent caselaw reading the same text the same way. Courthouse News covered the ruling the day it landed.
Read the Minnesota order past the headline
The part of that order worth an operator's attention is not the win. It is the caveat the court wrote into its own reasoning, in a passage that a press release will never quote:
"Plaintiffs have not shown that every event contract listed on Kalshi and Polymarket US fits the statutory definition of a 'swap.' If they don't fit, Plaintiffs have much weaker claims that the CFTC is the only authority that can regulate them."
The court went on to say that exclusive federal jurisdiction reaches transactions involving swaps conducted on DCMs, not every conceivable event contract a DCM might host, and that if the platforms are listing at least some contracts outside the swap definition, then any permanent relief may be much narrower than the preliminary order. That is a court, in the act of ruling for the industry, saying out loud that the industry's protection is contract by contract rather than company wide. It is the same conclusion we reached from the other direction when the CFTC's proposed rule finally defined its terms, which we wrote up in The Gaming Definition Is a Measuring Stick, Not a Threat.
Where the disagreement actually sits
It is tempting to file this as two judges with different instincts about gambling. The texts say otherwise. On 4 August 2026, in KalshiEX LLC v. Cox (No. 2:26-cv-00151-RJS-CMR), Judge Robert J. Shelby went further than New York had and granted Utah summary judgment outright, writing that "the court concludes the federal law relied upon by Kalshi does not preempt Utah's ability to enforce its anti-gambling laws." The memorandum decision is public on govinfo, and its express preemption analysis is the detail that should worry anyone building a compliance argument on intuition.
The CEA contains a provision, 7 U.S.C. section 16(e)(2), that expressly supersedes state laws prohibiting or regulating gaming. Read quickly, that sounds like the whole case. Read closely, as the Utah court read it, the clause reaches electronic trading facilities excluded under section 2(e) and certain excluded or exempted agreements, which is to say it preempts state gaming law for particular swaps that occur away from a designated contract market. The one express preemption clause in the statute points somewhere other than where the platforms need it to point. Whether that reading survives appeal is an open question in several circuits. The point for an operator is that the disagreement is about statutory text, not mood, and text based splits do not resolve quickly.
The map did not settle. It multiplied.
What followed the two rulings was not convergence. On 31 July 2026 New York Attorney General Letitia James and Governor Kathy Hochul announced a state court suit against Kalshi for running what the complaint calls an illegal, unlicensed gambling operation, seeking disgorgement, restitution and penalties at three times the gains, alongside a motion for a temporary restraining order.
On 11 August 2026 the CFTC answered with a tool that is not a court filing at all. Acting under section 8a(9) of the CEA, the Commission issued an emergency order directing Kalshi to keep operating. The order itself is worth reading for how the Commission frames the stakes: if New York's suit and the relief it seeks proceed, then in the Commission's words "a single State will effectively become the nationwide regulator of event-contract swaps on DCMs," which it calls the antithesis of the structure Congress designed. The order directs that Kalshi "shall continue to perform its functions as an exchange in accordance with the CEA's Core Principles and its normal practices."
Two days later a state court moved in the opposite direction. On 13 August 2026 the Washington Attorney General announced an order from King County Superior Court requiring Kalshi to stop offering contracts on sports, elections and politics, entertainment, culture, technology and science, and on whether public figures say particular words, with a first geofence due 19 August 2026 and a full multi source solution by 2 September 2026. As of 18 August 2026 that order stands.
Washington drew its line through the book, not around the company
The Washington order is the most instructive document in the whole sequence, and it has been reported mostly as a loss. Look at what it does. It does not expel the exchange from the state. It sorts the exchange's own catalogue and blocks part of it. Reporting on the order describes commodities, climate, economics and finance contracts as remaining available in Washington while the sports, politics, entertainment and culture categories go dark.
That is a court performing, from the outside and under deadline, exactly the exercise we have argued platforms should perform on themselves. A contract on a rate decision and a contract on whether a named athlete scores are not the same product, do not carry the same integrity risk, and were never going to receive the same treatment forever. Minnesota said so in a footnote to a ruling in the industry's favour. Washington said so with a geofencing deadline and a daily penalty attached. The difference between those two experiences is who did the sorting and how much time they had.
Why "wait for clarity" is not an operating plan
Our manifest puts the rule plainly in Directive 05: "Standards are set against the jurisdiction that regulates next, not only the one that has not yet." A contradictory map is the clean case for that directive, because it removes the usual excuse. When no authority has spoken you can argue about what is likely. When two federal courts have spoken in opposite directions, both readings are live, and only one of them is survivable if you guessed wrong.
So the operating rule while a split is open is the boring one: behave as though the stricter reading applies everywhere, and treat every favourable ruling as a preliminary posture rather than a permission. This is not pessimism about the category. We think prediction markets should exist and should grow. It is arithmetic about asymmetry. Operating under the looser reading and being wrong means disgorgement claims, geofencing under a court's clock, and users who discover mid position that their venue has changed shape. Operating under the stricter reading and being wrong means you left some revenue on the table for a year.
What the stricter reading looks like in practice
Concretely, four things, none of which require a final ruling to start.
- Classify your own book by contract, not by company. Both the Minnesota caveat and the Washington order sort contracts into groups. If you cannot produce, in an afternoon, a list of which of your markets rest on economic or commercial consequences and which rest on the outcome of a game, someone else will produce it for you and you will not like their categories.
- Build the geofence before a court names a date. Washington gave Kalshi six days to stand up a preliminary IP and residency filter and twenty days for the full version. Jurisdiction controls built calmly are a product feature. Built to a deadline with penalties running, they are an incident.
- Put the jurisdiction picture where the trade happens. Directive 02 asks for settlement sources, fees and conflicts to be stated in plain language on the surface where people trade. Access status belongs in that list while the map is unstable. A user in a contested state deserves to know their access rests on a preliminary injunction, and that is a sentence, not a legal memo.
- Write the open position policy now. More on this below, because it is the one nobody drafts until it is needed.
The part nobody plans: open positions when access disappears
Buried in a footnote of the CFTC's August order is a reference to an earlier one from 14 July 2026, staying an emergency rule Kalshi had filed and directing the exchange to fulfil open trades involving Michigan residents. Strip out the procedural language and the substance is this: when a venue is pushed out of a state, somebody still holds positions that were opened lawfully, in the venue's own account of things, and those positions have to be settled rather than stranded.
That question arrives before any appellate court answers the preemption question, and it arrives for the user rather than the lawyer. It belongs in published resolution terms before it is topical, which is Directive 01 doing ordinary work. We made the broader version of this argument in Venue Risk: You Are Not Only Trading the Event, and the split has since turned it from a thought experiment into a scheduling problem.
The bottom line
As of 18 August 2026 there is no single answer to whether a prediction market may operate in the United States, and nobody should write one. There is a federal regulator asserting exclusive jurisdiction and using emergency powers to hold a venue open, a proposed rule that has not been finalised, at least three federal district courts reading the same statute differently, an appeal pending in the Second Circuit, and a state court ordering a category by category shutdown with a deadline measured in days.
A category cannot plan around that by picking the ruling it likes. It can plan around it by assuming the strictest live reading is the one that will be tested against it, and by getting there before someone with subpoena power arrives. Our manifest says it in one line: the market that survives regulation is the one that was already behaving as if regulation had arrived. If your firm wants that to be a commitment rather than an intention, the signatory register is where it becomes public.