In January 2026, users who held correct NFL bets on Kalshi were only paid back their original stake, not their winnings. In February, a video editor for YouTuber MrBeast was suspended for suspected insider trading. In April, three congressional candidates were fined for betting on their own political races. In May, Kalshi froze $54 million in trades on the death of Iran's Supreme Leader.
2026 has been the year prediction markets got expensive lessons in what they can and can't be trusted to handle.
Here's the full case list, what each incident reveals about the industry's growing pains, and what it means for how you should approach these platforms now.
Case 1: The Kalshi NFL refund debacle
What happened: In January 2026, certain NFL player prop markets on Kalshi resolved with disputed outcomes. Users who held winning positions were initially paid back only their original stake — not the additional winnings they were owed based on the contract's stated payout structure.
The community response: gambling industry analyst Dustin Gouker publicized the issue. Coverage spread on prediction market Twitter. Within 48 hours, Kalshi reversed its decision and paid out full winnings.
What it revealed: Kalshi's expert-panel oracle has extremely broad discretion. Even markets with clearly stated resolution methodology can be overridden by internal review. There's no smart contract enforcement — Kalshi's willingness to pay out is what matters.
The lesson: read the contract terms twice. Then trade small first. If Kalshi's willing to reinterpret resolution rules on you, no amount of contract language protects you unless you can also organize public backlash.
Case 2: The MrBeast video editor insider case
What happened: In February 2026, Kalshi fined and suspended a video editor for MrBeast's YouTube channel. The specific allegation: the editor traded on markets related to MrBeast's video release timing and viewership numbers, using inside information about production schedules that wasn't public.
What it revealed: prediction markets have a genuine insider trading problem. Employees of companies with tradeable market questions have knowledge markets can't see. The current regulatory framework — CFTC oversight for Kalshi, no oversight for Polymarket — doesn't clearly criminalize this behavior the way securities laws do for public company insiders.
The lesson: if you're trading markets on outcomes tied to specific companies, individuals, or events with insider communities, assume you're not the smart money. Someone else has the information you don't.
Case 3: The congressional candidates betting on themselves
What happened: In April 2026, Kalshi fined and suspended three congressional candidates who had placed bets on their own political races. Each had traded YES on markets asking whether they'd win their primary or general election.
What it revealed: this isn't obviously illegal — but it clearly represents a conflict of interest and information asymmetry. The candidates had non-public information about their internal polling, fundraising, opposition research, and strategic decisions that would affect the outcome. Trading on that is at minimum an ethics problem and possibly a legal one under future prediction market regulation.
The lesson: watch who's trading. Some markets are structurally corrupted by insider participation. Political markets especially. If a candidate can bet on their own race, that market is not a clean signal.
Case 4: The Iranian Supreme Leader death market freeze
What happened: In May 2026, following Israeli airstrikes that killed Iran's Supreme Leader Ali Khamenei, Kalshi froze $54 million in trades on a market predicting whether Khamenei would be "out as Supreme Leader" by specific future dates. Kalshi stated the platform "doesn't allow transactions directly tied to death" and refused to resolve the market.
What it revealed: platforms can and do intervene when resolution is politically uncomfortable, even when contract terms clearly indicate a YES outcome. Users had trusted the resolution methodology; the methodology was overridden.
The lesson: prediction markets have discretionary limits that aren't fully visible until they're triggered. Markets that touch death, humanitarian crises, or specific political interventions can be frozen after the fact. Your position is only as safe as the platform's willingness to honor it.
Case 5: The Arizona criminal charges against Kalshi
What happened: In March 2026, Arizona Attorney General Kris Mayes filed 20 criminal misdemeanor charges against Kalshi, alleging illegal gambling and election wagering operations. The charges are pending. Kalshi called them "meritless" and is contesting.
What it revealed: even CFTC-regulated prediction markets face state-level legal exposure. Nevada issued a 14-day restraining order in March 2026. Massachusetts filed suit in September 2025. Each of these actions creates uncertainty about which markets can operate in which US states.
The lesson: regulatory patchwork is real. A market that's legal at the federal level may still be prohibited in specific states. Users in those states can find themselves participating in markets that are legally exposed in ways they didn't anticipate.
The pattern across all five cases
Each incident is distinct, but a common thread runs through them.
Discretion overrides contract. Platforms can and do choose to reinterpret resolution rules, freeze markets, or refuse payouts. The smart contract or contract language you read at trade time is not necessarily what governs your outcome.
Information asymmetry is not adequately policed. Insiders can trade in ways securities markets have spent 90 years learning to prevent. Prediction market frameworks are 5 years old at most, and enforcement is spotty.
Political sensitivity trumps market design. When a market touches politically or morally uncomfortable outcomes, platforms have shown willingness to intervene against clear resolution paths.
Regulatory environment is fluid. Federal approval doesn't equal state approval. What's legal today may be prohibited tomorrow. Position exit risk is not just market risk — it's regulatory risk.
What this means for how you should trade
The industry is growing up, and it's growing up messy. Five adjustments to how you approach prediction markets in 2026.
Size positions smaller. If a $54M market can be frozen, your $500 position is not guaranteed. Small positions across many markets diversify platform risk, not just outcome risk.
Diversify across platforms. Don't put your entire prediction market capital on one platform. UMA, Chainlink, and expert-panel oracles have different failure modes. Balance exposure across them.
Read platform terms of service. Especially the sections on market cancellation, freeze conditions, and dispute resolution. If a platform reserves the right to freeze markets, they will exercise that right eventually.
Avoid politically sensitive markets in size. Elections, geopolitical outcomes, and death-related contracts have been shown to attract discretionary intervention. Fine for small positions, dangerous for large ones.
Watch the news for regulatory shifts. State-level actions against Kalshi, Thailand's Polymarket block, Australia's full ban — each creates position exit issues for users in those jurisdictions. Stay informed on the regulatory environment where your funds sit.
What Thai users should specifically know
Thai users considering prediction markets in 2026 face compounded risks compared to US or EU users.
No Thai regulatory recourse. If Polymarket freezes a market or Kalshi reinterprets a resolution, there's no Thai regulator to appeal to. You're relying entirely on offshore platform good faith.
No consumer protection framework. Thailand hasn't developed a prediction market consumer protection regime. Users have no assurance of fund recovery if a platform fails or blocks their positions.
Cross-border tax exposure. Any gains from Polymarket (however accessed) are taxable in Thailand. Losses may not be deductible. This asymmetry is unfavorable.
Enforcement uncertainty. The TCSD's January 2025 Polymarket block sets a precedent. Other platforms could be blocked at any time, with existing user positions potentially locked.
The trust question that isn't going away
Prediction markets are built on trust — trust that the platform will resolve markets correctly, trust that oracles will report truthfully, trust that insiders can't game the system. When any of those trust assumptions fail, the whole platform's value proposition weakens.
2026 has been a stress test for that trust. Kalshi has faced multiple challenges. Polymarket has faced regulatory blocks in growing regions. New entrants like World are trying to solve the trust problem with different technical approaches. None has fully cracked it.
What all of this points to is the case for locally-regulated, jurisdiction-aware prediction markets. Not because they're better than global platforms in every way, but because they solve trust differently — through recognizable legal recourse, transparent platform accountability, and audit trails that survive contact with regulators.
What Juno is doing about this
Juno's operating principle: assume every discretionary intervention will happen at some point, and build the platform to minimize their impact.
Contract terms are published and unchangeable once markets open. No mid-market rule reinterpretation. Platforms that reserve the right to override contract terms have shown that right will be exercised.
Resolution sources are pre-specified with fallback procedures. If the primary data source is unavailable, a secondary is triggered automatically. No expert discretion needed for the vast majority of cases.
Insider trading detection is built in. Trading patterns that suggest insider information trigger review before payout. This isn't perfect — no system is — but it's dramatically better than "we didn't check."
Local regulatory recourse is real. Because Juno operates within Thai regulatory frameworks, Thai users have appeal mechanisms if platform behavior deviates from stated terms.
The five-case lesson
Prediction markets are legitimately powerful. They also have real, demonstrated failure modes. 2026 has been the year those failure modes became visible.
If you're trading these platforms, you can't just read the headlines about how markets beat polls in 2024. You have to also read about the NFL refund, the MrBeast editor, the Iranian death market freeze, and the Arizona charges. Those are the receipts on the other side of the ledger.
Prediction markets will still beat traditional forecasting on most questions. But the platforms running those markets are not infallible. Trade accordingly.