There is a specific kind of fraud that prediction markets make possible that nothing else quite does. If you know something important before the public does, you can bet on it and collect. That is not a theoretical risk. It has already happened multiple times, openly, and with very large sums.
A US Army Special Forces sergeant was charged in April with using classified details about a planned operation to arrest Venezuela's Nicolás Maduro. He opened a Polymarket account on December 26, 2025, placed 13 bets totalling around $34,000, and walked away with over $400,000 after the arrest went ahead on January 3. He then tried to get his account deleted and changed the email tied to his crypto wallet. The Department of Justice and the CFTC both filed charges.
That case was the clearest, but far from the only one. A New York Times investigation found over 80 accounts with suspicious betting patterns on Polymarket, spanning 30 separate topics from Israeli military strikes to regulatory decisions on financial products. One trader made nearly $1 million with a 93% success rate on bets tied to unannounced US and Israeli military operations against Iran, placing those bets hours before each strike across three separate incidents. At least 50 freshly created accounts placed bets on a US-Iran ceasefire on April 7, some of them opened minutes before the announcement became public.
Israeli authorities separately indicted two individuals, including a military reservist, for using classified information to bet on Iran-related outcomes on Polymarket.
The market that all of this is happening inside has grown at an extraordinary pace. Monthly trading volume across Kalshi and Polymarket rose from roughly $2 billion in early 2025 to a peak of nearly $27 billion in January 2026. Those two platforms now control about 97% of all prediction market activity globally. Investment firm Bernstein forecasts the sector could reach $1 trillion in annual volume by 2030.
The regulator watching over all of this, the CFTC, has simultaneously shrunk. Its headcount dropped from around 708 staff to roughly 543 over the past year, a cut of more than 20%, even as Congress prepares to give the agency more responsibility over crypto markets. Chairman Michael Selig has been direct about how he plans to compensate: AI. The agency is training all staff on Microsoft's Copilot, building internal tools to scan trading data for suspicious patterns, and using those tools to decide when to open investigations or send formal demands for trader information.
The mechanics of Polymarket make AI surveillance more tractable than you might expect. All trades on its platform are recorded publicly on the blockchain, a permanent and openly readable ledger. That means regulators and academics alike can reconstruct betting activity with second-by-second accuracy. The challenge is not finding the data, it is knowing which patterns actually indicate insider knowledge rather than a lucky streak or a sophisticated betting strategy.
Polymarket has also brought in the blockchain analytics firm Chainalysis to help monitor activity. Kalshi has updated its rules to block politicians, athletes, and other conflict-prone individuals from trading in certain markets. The US Senate voted unanimously to ban its members and staff from trading on prediction markets entirely.
But the structural problem runs deeper than any single enforcement action can fix. Polymarket's main platform remains technically offshore, and US users who want to access it have been using tools that disguise their internet location to bypass the block. The CFTC says it can reach those users regardless. It filed its own complaint against the Venezuela soldier even though the platform he used is not registered with US regulators.
The deeper tension is this: the same administration that cut the CFTC's workforce is also the one that opened up prediction markets to far more activity than ever before, withdrawing previous rules that restricted what could be listed for trading. The market got bigger and the watchdog got smaller at the same time. AI surveillance is being asked to close that gap. That is a bet worth watching.