Is Kalshi Gambling? What the CFTC and the Trademark Filing Both Say

The regulatory answer

On 4 November 2020 the CFTC issued an Order of Designation to KalshiEX LLC, granting it status as a designated contract market, or DCM. The order came under Section 5 of the Commodity Exchange Act and CFTC Regulation 38.3(a), and the Commission's stated basis was that KalshiEX had demonstrated its ability to comply with the Act and the regulations applicable to DCMs.

That is the whole content of "Kalshi is CFTC-regulated", and it is narrower than the phrase suggests. It is a finding that an entity can run an exchange: not a licence to list any particular contract, not a finding that any contract is lawful in any particular state, and not a consumer-protection regime of the kind a gaming regulator operates.

Read directly at CFTC Release 8302-20 on 11 August 2026. The designation order is on the same server.

What a designated contract market is

A DCM is a regulated exchange. It self-regulates against core principles: monitor its own market for manipulation, publish and enforce its rules, report to the Commission. The CME and ICE hold the same designation, so Kalshi is a small venue with an unusual product list rather than a different species of institution. Two things follow. Your trades clear through a registered clearing organisation, so you do not depend on the other side to pay when you win. And the designation says nothing about whether a contract is a good trade, or lawful where you live.

The boundary that decides everything else is one word: swap. The CFTC's exclusive jurisdiction runs to swaps traded on a DCM. If a sports event contract is a swap, federal jurisdiction is exclusive and state gambling law is displaced. If it is not, the state keeps the authority over wagering it has always had. Every state fight below turns on that question, and courts have now answered it both ways.

The trademark filing that says otherwise

In early November 2025 Kalshi applied to the USPTO to register the term "prediction market". An application has to describe the services the mark would cover, because a registration is only ever as broad as the services claimed. Alongside the financial trading categories, according to Sportico's reporting of the filing, the description listed "bookmaking services, namely, providing of information related to sports betting; organizing, arranging, conducting sports betting and gambling tournaments, competitions and contests." That is not a journalist's characterisation of Kalshi. It is Kalshi's counsel, writing to a federal agency, in the one document whose whole purpose is to state what the business does.

Kalshi's answer came through a spokesperson, Elisabeth Diana: "This is not a characterization of our business as anything other than prediction markets." The company called the breadth defensive, a way of stopping others blurring the distinctions between products. Take that seriously, because applications really are drafted wide to fence off adjacent uses. It is a genuine explanation and an incomplete one: a wide net still gets cast over water you expect to fish in.

Then two things happened that are more informative than the original wording. The USPTO refused the application, holding that "prediction market" merely describes a feature of the services rather than identifying their source. And after Sportico reported the discrepancy, Kalshi filed a fresh batch of applications with the gambling and bookmaking language removed, reframed in financial-exchange terms. So its position is not that the original wording was misread. Its position, revealed by conduct, is that the wording was a liability.

Why both are true at once

The two answers do not collide, because they answer two different questions. Is Kalshi regulated as gambling? No. It is regulated as an exchange, by a federal financial regulator, under a statute about commodities. Is what happens on Kalshi gambling? That depends on the contract, and on the sports contracts courts have started saying yes.

On 10 August 2026 Judge Vernon D. Oliver of the District of Connecticut denied Kalshi a preliminary injunction against the state's regulator, on the definitional point above: sports event contracts do not satisfy the statutory definition of a swap, and even if they did, Connecticut's gambling laws would not be displaced. The ruling was reported to have observed that Kalshi itself has promoted the platform as offering legal sports betting nationwide.

Two weeks earlier, on 27 July 2026, Judge Katherine Menendez of the District of Minnesota reached a partly opposite conclusion, enjoining Minnesota from enforcing its new statute because the law was likely at least partially preempted, the CFTC having exclusive jurisdiction over swaps traded on DCMs. She then drew a line through the product list: that reasoning reaches contracts with clear potential economic, financial or commercial consequences, and may not reach a contract on a reality television result.

Read the two together and the answer falls out. It is not one answer for the company, it is a different answer per contract. A contract on a central bank decision looks like a swap to a court, because it references a variable real balance sheets are exposed to. A contract on Sunday's game does not. Same venue, same order book, same fee formula, two legal characters. Which makes the right response to "is Kalshi gambling" a question back: which market.

The distinction that actually matters to you

The taxonomy fight changes nothing about your arithmetic. Two other distinctions do.

Hedging versus speculation

The Commodity Exchange Act has never required you to be hedging. Futures markets are built on commercial participants passing price risk to speculators willing to take it on, so speculation is a feature of the design rather than a loophole in it. A farmer short corn is hedging, a trader long corn with no corn is speculating, both are lawful on a DCM, and the exchange does not ask which you are.

So the test that works is a sentence you can either finish or not: name the exposure this position offsets. If you can, you are hedging, and it should be sized against that exposure rather than against your appetite. If you cannot, you are speculating, which is legal, priced, and has an expected value you can calculate. A negative-expectancy position does not turn positive because a federal agency regulates the venue.

Where the label does bite is downstream. Gambling regulators run consumer-protection machinery the CFTC framework does not contain: self-exclusion registries, deposit limits, advertising rules aimed at problem gambling. Tax treatment differs between derivative positions and gambling winnings. And a state regulator can switch a venue off while you hold a position, which has fired repeatedly through 2026.

What the fee structure implies about the venue's view

This is the part nobody answering the question has worked through, and the most informative thing available about how Kalshi understands its own product. The taker fee, from its published schedule:

C is the number of contracts and P is the price in dollars. Because C sits inside the ceiling, the round-up applies to the whole trade rather than per contract. Makers pay a quarter of that rate, 0.0175. S&P 500 and Nasdaq-100 tickers are charged at half the taker rate, 0.035.

Now look at what P x (1 - P) is. For a contract that settles at one dollar or at nothing, P(1 - P) is exactly the variance of the outcome. Kalshi is not charging a percentage of your stake, and not a percentage of your notional. It is charging a fixed multiple of the variance of the thing you are trading.

The fee in dollars is symmetric about fifty cents and peaks there. As a share of what you put up it is not symmetric at all, falling from roughly 6.8 per cent of stake at a nickel to four tenths of one per cent at ninety-five cents. Because contracts equal stake divided by price, the fee comes to stake times rate times (1 - P), so break-even as a fraction of stake is just rate x (1 - P).

Compare that with how a sportsbook earns. A book's margin is the vigorish, built into the price it publishes, present across the board and fattest where the customer is least informed: longshots, exotic props, parlays. Long-shot bias is among the most durable findings in the literature on betting markets, so a house pricing for action would price hardest at the tails. Kalshi prices almost nothing there: at three cents its taker fee is under a quarter of a cent per contract.

Charging on variance is charging for uncertainty. Fifty cents is where the market does not know, where the matching service is worth most, and where a maker holds the most inventory risk. Ninety-seven cents is nearly resolved, carries almost no information, and is charged accordingly. The maker side says it louder: makers pay a quarter of the taker rate here, and on Polymarket a verified zero, where the taker formula is C x rate x P x (1 - P) with category rates from zero to seven per cent. A bookmaker never rewards you for posting a price, because the price is the product it sells.

Do not overread it. Variance-proportional pricing on a bounded contract is also just sensible inventory-risk pricing any derivatives venue would reach independently, and roulette has variance too. The half rate on index tickers cuts both ways: the venue's own taxonomy separates index futures from everything else on the board. A fee schedule tells you how a venue models its product, not what customers do with it. But if you want to know what a business thinks it is selling, read its price list before its press releases, and Kalshi's is priced on doubt.

What the states are doing about it

Federal designation has settled nothing at state level. Thirty-eight state attorneys general plus the District of Columbia signed an amicus brief supporting Massachusetts against Kalshi, released on 25 April 2026 in Suffolk County Superior Court. The argument was jurisdictional, not moral: Kalshi's "aggressive theory of preemption threatens the States' longstanding ability to protect their citizens in this area."

The tribal-gaming objections

This is the least covered part of the story and structurally the hardest for Kalshi, because the objectors are neither states nor the federal government. The Indian Gaming Regulatory Act, 25 U.S.C. 2701, makes Class III gaming, which includes sports betting, lawful on tribal lands only under a tribal-state compact. Tribes bought that exclusivity with revenue sharing and a heavy regulatory burden, so a sports event contract sold into tribal territory by a federally designated exchange is, on their argument, an end-run around the bargain.

Four New Mexico tribes, the Mescalero Apache Tribe, the Pueblo of Isleta, the Pueblo of Pojoaque and the Pueblo of Sandia, sued Kalshi on 13 May 2026 alleging illegal sports betting on tribal lands in violation of IGRA. It was reported as the third tribal action, after California and Wisconsin. In the Wisconsin case a federal judge declined the Ho-Chunk Nation an immediate injunction while finding the tribe had shown a likelihood of success on its IGRA claims, which is a warning rather than a win. The California case reached the Ninth Circuit, argued 13 July 2026. No preemption argument resolves this cleanly, because here it sets one federal statute against another. That is why the tribal track is the one to watch.

What we can show from our own record

Everything above this heading is documents. The empirical answer to the headline question is calibration, we do not have it yet, and so this section is a method rather than a finding and carries no figure.

Calibration means this. Take every Kalshi market that traded in a given price band, say seventy to eighty cents, wait for them all to resolve, and count the fraction that resolved YES. If the market is well calibrated that fraction lands inside the same band: contracts priced at roughly three quarters come in roughly three quarters of the time. Do it for every band and you have a reliability curve for the venue.

Why that settles the argument instead of continuing it: a roulette wheel cannot be calibrated. Its prices carry no information beyond a fixed house edge, and no volume of data will make them informative, because there is nothing there to be informed about. A forecasting instrument is precisely one whose prices track realised frequencies. If Kalshi's are calibrated, its defenders are right about what the venue is for. If they are systematically off, the prices are a game surface with a spread on it. Either way it is a measurement anybody can check, and it needs no agreement about what "gambling" means. Here is the method, written down before the numbers exist so it cannot be chosen to flatter them.

1. One price per market, at a fixed horizon before close, timestamped. A contract polled four hundred times is still one observation of one outcome, and a long-lived market would otherwise dominate its band.
2. Ten price bands, set once in advance and never redrawn to improve a curve.
3. Resolution from the venue's own settlement record, not from a news report of the event.
4. Every band publishes its YES rate, its market count and an interval. The count sits beside the rate every time, because a band holding twelve markets is not evidence.
5. Selection stated, not hidden. Long-dated contracts drop out of an early sample, so a first publication is a sample of short-horizon markets and says so.
6. A Brier score and a reliability diagram alongside the bands, misses included. A curve showing only the bands that behaved is not a reliability curve.

There is no placeholder standing in for the figure. Calibration needs a full resolution cycle to close before the first band means anything, which is eight to twelve weeks at minimum. Publishing early would be publishing noise with a decimal point on it, and a decimal point is what makes noise look like evidence.

One further admission, since this section is about our own record. Our paper trading scanner is not currently running: it crashes on start under Windows, on a module that exists only on POSIX systems. Until that is fixed and a cycle has closed, no figure derived from it appears anywhere on this site, and any figure you may have seen from it previously should be treated as withdrawn.

Sources, and what we actually read

Citing a source is table stakes. Saying which we read ourselves and which reached us second hand is not, so every claim is marked with how we got it. Where that is unflattering, it is still the answer.

Disclosure