Polymarket promises to police itself. A billion-dollar U.S. franchise rides on it

Two months before the U.S. midterms, Polymarket told Reuters its surveillance systems are ready. A new global head of investigations who is a former FBI investigator, machine-learning and blockchain analytics, a public transparency page, more than a hundred cases already referred to law enforcement. On its face that reads as a compliance footnote. It is not. The announcement is aimed less at rule-breaking bettors than at the people who decide whether the world's biggest prediction market gets to keep operating in the United States — a permission now worth real money.

Here is what that permission has become worth. Polymarket trades contracts that pay a dollar if some event happens — the Democrats sweep Congress, a team wins a final — and nothing if it doesn't. The price is the market's implied probability. For most of its life the platform charged nothing, and it was a curiosity: about $73 million of total trading volume in all of 2023. Then the 2024 election arrived, the federal legal wall came down, and in 2026 Polymarket switched on trading fees — a royalty measured in basis points, not percent. By late June, Reuters reported the company had topped an annualized $1 billion in revenue, up from effectively zero in 2025, when it charged no fees at all. Monthly volume set a record near $10.5 billion in March 2026, and politics is roughly a third of the flow.

The platform never takes the other side of a bet; it takes no proprietary positions and warehouses no risk. It is a toll booth on other people's gambling. That is the business Intercontinental Exchange, the parent of the New York Stock Exchange and a company with roughly $10.5 billion in trailing revenue, decided it wanted: ICE has put about $2 billion into Polymarket for roughly a fifth of the company at an $8 billion valuation, framing the deal as financial data infrastructure, not wagering.

Now the part a surveillance promise cannot fix. A prediction market works by rewarding people for acting on information before the crowd does — anonymous, cash-backed, immediate. The more exclusive the knowledge, the better the trade. That is not a bug leaking in from the edges; it is the core mechanism, turned up as high as security allows. Insider trading is the product's own engine running at full speed, and the cases are already public. A U.S. Army master sergeant is accused of wagering roughly $33,000 across 13 contracts on Nicolás Maduro's capture in the week before the special-operations raid, using classified information, then cashing out for more than $400,000. On an April day this year, at least 50 freshly created anonymous wallets bet on a U.S.–Iran ceasefire in the hours and minutes before President Trump announced one; one wallet opened 12 minutes prior and cleared about $48,500, with accounts making roughly $1 million in aggregate on the strike-date question. Polymarket says it has referred more than 100 cases to law enforcement, and the CFTC is reported to be running about 20 investigations built on evidence from rival Kalshi alone.

The structural problem is sharper than the crime stories, because the numbers the franchise and its valuation are built on are the hardest ones for a toll-booth operator to police honestly. Columbia researchers found that roughly a quarter of Polymarket's trading volume over three years looked like wash trading — wallets trading against themselves to manufacture activity — peaking near 60% of weekly activity in December 2024, across networks estimated at tens of thousands of accounts. Fees scale with volume. So the operator is paid on the meter while being asked to catch the meters it profits from. Self-policing has a poor historical record: the CFTC was created in 1974 in part because exchange self-regulation had demonstrably failed, and the agency has spent the decades since prosecuting spoofing and wash trading on registered futures and equities markets that at least have public order books and identified parties. Polymarket's version is harder than all of that — pseudonymous wallets, a public blockchain ledger, and a geofence that on-chain researchers say many U.S. users slip around, estimating U.S.-linked wallets traded roughly $571 million of political contracts in a year as the largest national group despite the ban.

Which raises the question of why Polymarket would volunteer any of this now. Because the permission structure is being rewritten in real time, and the midterms are the exam. For more than a decade the CFTC's official position was that election contracts were contrary to the public interest; in January 2022 it fined Polymarket $1.4 million for running an unregistered market and ordered it to block U.S. customers. A federal judge threw out the agency's ban on Kalshi's election contracts in September 2024, an appeals court declined to freeze the decision, and election markets went live weeks before that year's vote. Polymarket bought its way back in — acquiring a CFTC-licensed exchange and winning staff no-action relief on event-contract reporting in 2025 — and the commission now has a chairman arguing publicly that states should not muscle in. The CFTC opened its own rulemaking in March 2026 and proposed a public-interest framework in June. Yet the same Washington that legalized the category is frightened of it: the Senate voted unanimously in April to bar its own members and staff from prediction markets, Democrats have pressed the agency to rein the industry in, states are suing operators as unlicensed gambling, and CNBC reported this summer that the CFTC had an extensive, ongoing investigation into Polymarket itself — even as the company says a separate probe into its old settlement was dropped. The "surveillance ready" statement is pre-committed evidence for that audience: we can police this category, so you do not need to shut it down.

If your interest is investing in the boom, the first fact is an anticlimax: Polymarket is private and has no coin. The one clean public-company window is ICE, and its seat is a sideshow relative to the parent — a $2 billion stake small against an exchange-and-data company booking more than $10 billion of annual revenue — whose stated thesis is data, not a cut of bets. The structurally interesting money lands elsewhere. As event contracts become CFTC-covered derivatives, settlement has to run through licensed, trusted data suppliers; that is the logic that plants sports-data providers with audited financials, like Genius Sports and Sportradar, at the center of regulated event markets while the betting platforms stay private. The asymmetry is the point: the casino is private and a policy question, but the roulette wheel's data is public.

So watch the odds least of all. Watch the two-month test. If the CFTC's final rule stays permissive and the midterms produce no fresh scandal on the scale of a soldier betting on a classified raid, the franchise keeps its politics category — a third of its volume — and the billion-dollar run-rate business stays legal. If rule-writers instead conclude that a pseudonymous market cannot police the very behavior it pays for, the U.S. political book closes, and you find out how much of that volume was franchise and how much was a two-year anomaly. Prediction markets exist to let people profit from knowing things before others do. Whether the U.S. lets that business run comes down, in the next two months, to whether a toll booth can prove it polices the behavior rather than the information.