Prediction market disputes: who decides, and the fight over the rules

The short version

• A prediction market dispute is one of four things: a trader against the platform, a fight over how a market resolved, market abuse such as insider trading, or the federal and state fight over whether these markets are gambling. This guide maps all four and hands the detail to four companion pieces.
• US prediction markets run as CFTC-registered designated contract markets, or DCMs, a kind of derivatives exchange. The Commodity Exchange Act gives the Commodity Futures Trading Commission (CFTC) “exclusive jurisdiction” over their contracts (7 U.S.C. § 2(a)(1)(A)).
• How a market resolves depends on the venue. Kalshi, a regulated exchange, settles centrally under its own rulebook against sources it names before trading; Polymarket settles through the UMA optimistic oracle, where a proposed outcome stands unless someone challenges it within two hours and a token-holder vote decides.
• Disputes are rare but concentrated. One study of 802,863 resolved Polymarket markets found that, among the 529,408 with an observable dispute log, 0.31%, about one in 320, were formally challenged, and those markets carried a far larger share of trading volume.
• Who may regulate is unsettled. The Third Circuit held the CEA preempts state gambling law (April 2026); the Ninth Circuit found no likely preemption for sports event contracts (August 2026); the Supreme Court has the question but has not taken it up (No. 26-299, as of 30 October 2026).^{1,13,14,15,20,21,22}

What is a prediction market dispute?

A different body decides each of the four. A prediction market is an exchange where people trade event contracts, contracts that pay a fixed amount, usually $1, if a stated event happens and nothing if it does not. When something goes wrong the question is always the same: who decides, and under what rule.¹⁶

A dispute with the platform, over an account, a payout or a fee, is handled by the exchange’s own rulebook and, where a registered intermediary is involved, by CFTC reparations: see how a user-versus-platform dispute is resolved. How a market resolved, the outcome itself, is decided by the platform’s resolution process, a rulebook decision on Kalshi or an oracle and a vote on Polymarket: see who decides an outcome and how to challenge it. Market abuse, including insider trading, is handled by the platform’s rules and the CFTC: see prediction market insider trading. Whether these markets are even lawful is for the courts, the CFTC and the states: see the lawsuits. This guide covers the United States. For dispute resolution, online dispute resolution and ombudsman schemes generally, see AI in dispute resolution; for complaints and early resolution, AI in dispute avoidance.

Are prediction markets gambling or regulated financial contracts?

In the United States, prediction markets are event contracts listed on CFTC-registered exchanges, not casino bets, at least as a matter of federal registration. The CFTC treats a qualifying event contract as a kind of swap, a derivative whose value turns on whether an event happens, and claims exclusive jurisdiction over it. That claim is exactly what the states dispute, and the tension runs through the whole sector.^1,16

• The grant of “exclusive jurisdiction” over swaps and futures on registered entities sits in the Commodity Exchange Act (CEA) at 7 U.S.C. § 2(a)(1)(A).
• A registered prediction market operates as a designated contract market (DCM). Kalshi (KalshiEX LLC) was designated on 3 November 2020.
• Polymarket’s US entity, QCX LLC (Polymarket US), was designated a DCM on 9 July 2025, with an amended order following petitions submitted through 24 November 2025. Its international exchange is not US-domiciled and says it blocks US users.
• DCMs list event contracts by self-certification, with no prior CFTC approval: the exchange certifies that the contract complies with the Act (17 CFR § 40.2; 7 U.S.C. § 7a-2(c)).
• The Special Rule lets the CFTC find a contract “contrary to the public interest” if it involves unlawful activity, terrorism, assassination, war or gaming (7 U.S.C. § 7a-2(c)(5)(C); 17 CFR § 40.11). The Congressional Research Service reports that, under current leadership, the CFTC has not used the Special Rule to review a self-certified event contract.
• On 19 October 2026 the CFTC issued an interim final rule codifying that casino-style gambling products, including sportsbook and casino wagers, are excluded from the “swap” definition, the dividing line the sports-contract fight turns on (91 FR 65969).^{1,2,5,6,10,11,12,17}

The gambling-versus-derivatives argument sits at the centre of the court fight, and it is where the states and the CFTC collide: see prediction market lawsuits.

How does a dispute between a user and the platform get resolved?

Through the exchange’s own rulebook and its dispute-resolution facilities, not a public court by default. A trader’s recourse is narrow: their path to a federal regulator does not run against the exchange itself. A trader who was not paid, or who says the platform took their money, falls here, and the first forum, often the only one, is the exchange that wrote the rules.³

• A DCM “shall establish and enforce disciplinary procedures” (Core Principle 13) and “shall establish and enforce rules regarding, and provide facilities for alternative dispute resolution as appropriate for, market participants and any market intermediaries” (Core Principle 14): 7 U.S.C. § 7(d)(13) to (14), carried into 17 CFR Part 38 subparts N and O.
• Core Principle 14 requires ADR facilities. It does not itself order arbitration; what a trader must do, and whether a claim is heard on the papers or at a hearing, lives in each platform’s rulebook and user agreement.
• CFTC reparations are the federal route, but they run against a registered intermediary such as a broker, not against the exchange, and must be filed within two years (7 U.S.C. § 18(a); 17 CFR Part 12). A complainant picks a track by the size of the claim: a $50 voluntary track decided with no appeal, a $125 summary track for claims up to $30,000 decided on the papers, or a $250 formal track for larger claims that can carry a hearing.
• A consumer class action against Kalshi has been reported, alleging it violated state gambling laws and misled users about how the platform operates. Treat it as an allegation; we found no verified docket as of 30 October 2026.^{3,4,7,8,9,13,27}

Where to complain about an exchange or a scam is covered in crypto exchange complaints; arbitration of futures-industry disputes is at /research/nfa-arbitration/.

Who decides how a prediction market resolves, and can you challenge the outcome?

The two leading venues decide outcomes in opposite ways. Kalshi settles centrally under its own exchange rulebook, against sources named before trading starts, with the exchange making the determination, not a crowd. Polymarket decides by a decentralised oracle that anyone can challenge and that a token-weighted vote settles. Set side by side, Kalshi resolves through a named decider and Polymarket through a crowd.^13,14

On Polymarket, a proposed outcome goes to the UMA optimistic oracle, a smart-contract system that treats a proposal as correct unless it is challenged. The proposer posts a bond, stated as $750 on Polymarket’s help pages as last updated in January 2026; the outcome is final unless someone disputes it within a two-hour window; a disputer matches the bond; and UMA token holders then vote, after a debate of roughly 24 to 48 hours. Where an outcome is genuinely unresolvable, Polymarket’s help pages say UMA can settle the market as a 50/50 split. Once UMA finalises a market the result is immutable, and Polymarket says it never holds the funds itself (it is non-custodial) and cannot reverse the result. A large holder swaying that vote, an oracle or whale governance attack, is the market-manipulation worry on this side.^13,14

Disputes are uncommon but not evenly spread. In the study of 802,863 resolved markets, 1,634 of the 529,408 markets with an observable dispute log, 0.31% or about one in 320, were formally challenged, and disputed markets carried about 5.8% of all trading volume. Dispute rates vary sharply by resolution source and climb as the resolution wording gets harder to read; the full breakdown by source is in prediction market resolution disputes. Harder-to-read resolution criteria and an unclear source of truth go with more disputes.¹⁵

The wording decides. Someone whose position resolved No, and who wants to know why, is asking two things: what the rule was, and who applied it. Ambiguous wording and an unclear source of truth are what turn a settlement into a dispute. For the full challenge steps, the deadline and cost, and whether you get your stake back, see prediction market resolution disputes; for the oracle and the decentralised-finance code behind it, smart contract dispute resolution.¹⁵

What counts as insider trading or market abuse in a prediction market?

There is no blanket ban on trading with nonpublic information on a prediction market. Liability turns on misappropriation, trading on material nonpublic information (MNPI) in breach of a duty, under CFTC Rule 180.1, and on each platform’s own rulebook prohibition. Rule 180.1 is modelled on SEC Rule 10b-5 and is used for misappropriation cases; it does not set up a rule that everyone must trade on equal information.¹⁶

On 25 February 2026 the CFTC’s Division of Enforcement issued an advisory setting out its authority over conduct on a DCM: insider trading under CEA § 6(c)(1) and Regulation 180.1, wash and noncompetitive trades under § 4c(a), disruptive trading under § 4c(a)(5), and the exchange’s own duties under the Core Principles of § 5(d). The same advisory records two Kalshi penalties as examples. A political candidate who traded on his own race drew a $2,246.36 penalty and a 5-year suspension (May 2025); a video-channel editor who traded on nonpublic information drew a $20,397.58 penalty and a 2-year suspension (August and September 2025). Those were exchange penalties, not CFTC orders.^18,19

The CFTC has since opened its own enforcement actions over prediction-market trading; all contested matters are allegations until decided. The decentralised analogue is the whale governance attack on an oracle vote, described above. The framework, the venues and the disciplinary process are set out in prediction market insider trading.¹⁷

Who gets to regulate prediction markets, the states or the CFTC?

Who regulates prediction markets is unsettled, and it is the biggest open question in the sector. Whether prediction markets are legal can turn, state by state, on the answer. There is a genuine circuit split and a pending Supreme Court petition. The dispute is whether the CEA’s grant of exclusive jurisdiction to the CFTC gives it federal preemption over state gambling law for sports event contracts.¹

• Third Circuit: KalshiEX, LLC v. Flaherty, No. 25-1922 (3d Cir.), decided 6 April 2026, 2 to 1. The court held that “both field and conflict preemption apply,” so the CEA preempts New Jersey’s gambling law as applied to Kalshi’s sports event contracts.
• Ninth Circuit: KalshiEX, LLC v. Assad, No. 25-7516 (9th Cir.), decided 28 August 2026. The panel held that Kalshi “did not show a likelihood that the CEA preempts” state gaming rules, because the sports event contracts “were not ‘swaps.’” That disagreement with the Third Circuit is the circuit split.
• Supreme Court: Flaherty v. KalshiEX, LLC, No. 26-299, petition filed 2 September 2026 and docketed 8 September. As of 30 October 2026 the response is extended to 7 December 2026, and the Court has neither granted nor denied review.
• The CFTC’s counter-offensive: United States and CFTC v. State of Utah, No. 2:26-cv-01055 (D. Utah), filed 26 October 2026, asks for a declaration that state gambling law is “preempted and unenforceable as applied” to CFTC-regulated markets. The complaint lists parallel preemption suits against Arizona, Connecticut, Illinois, Kentucky, Minnesota, New Mexico, New York, Rhode Island and Wisconsin.
• The trigger in Utah was a civil suit: Brown v. KalshiEX, Utah 3d District Court No. 260908356, filed 20 October 2026, seeking a declaratory judgment, an injunction, restitution, disgorgement and punitive damages on gambling and public-nuisance theories. It is a civil action.^{20,21,22,23,24}

Congress has many pending bills on prediction markets; the inventory is in the Congressional Research Service report R49266. For who is suing whom, and the four litigation tracks kept apart, see prediction market lawsuits.¹⁷

What is settled, and what is still open?

Two things are reasonably settled; the central question is not.

• Settled: Polymarket’s unregistered-venue history. In re Blockratize, Inc., d/b/a Polymarket.com, CFTC Docket 22-09 (3 January 2022), brought a $1.4m penalty and a wind-down of non-compliant markets.
• Settled: Kalshi’s election-contract fight. The D.C. Circuit denied the CFTC’s request to stay the district court’s ruling for Kalshi (No. 24-5205, 2 October 2024); election contracts are now among the contracts listed on CFTC-regulated exchanges.
• Open: whether the CEA preempts state gambling law for sports event contracts (the split, with the Supreme Court petition pending), whether and how the Special Rule will be used, and the final shape of the 2026 CFTC rulemaking on event contracts.^{2,17,20,21,22,25,26}

What would make prediction market disputes fairer and faster?

In our view, every one of these disputes already runs on a written rulebook. A user agreement or an exchange rulebook: the terms exist before the trade. The disputes that do happen are rare but alike. The study of 802,863 markets found formal challenges in only 0.31% of markets, concentrated in high-volume contracts, and found that harder-to-read resolution criteria went with more disputes. What makes a dispute fair and fast is running each one the same way under those rules. A named person or body should decide, which is what Core Principle 14’s dispute-resolution facilities already assume (7 U.S.C. § 7(d)(14)). Any AI should read and draft, not decide. And there should be a record each side can check, and that a court or the CFTC can later rely on, because reparations and disciplinary procedures are built around a reviewable record (17 CFR Parts 12 and 38). The study backs the first half of this: clearer resolution wording went with fewer disputes. Whether one named decision beats a vote taken after the fact is not something the study tested; that preference is ours. It is an opinion, not a promise about any product.^3,7,9,15