Prediction Markets Are Exploding. So Are the Insider Trading Cases.
Prediction markets are real money platforms where users wager on real world outcomes such as elections and trending lists, priced like trades.
Prediction markets are real money platforms where users wager on real world outcomes such as elections and trending lists, priced like trades.
A prediction market is a real-money platform where users bet on the outcome of real-world events, from elections and corporate announcements to celebrity rankings and geopolitical events. Prices move like trades, so the going rate functions as a crowd-estimated probability. In May 2026 federal prosecutors charged a Google software engineer, Michele Spagnuolo, with turning that setup into an alleged insider-trading scheme: he wagered roughly $1.2 million on Polymarket, a real-money prediction-market platform, on which artists would top Google's year-end most-searched list, leveraging confidential company data he had access to in his role at Google (CNET; DOJ SDNY press release).
The Spagnuolo case is charged, not convicted, and turns on whether confidential access to internal Google data about search trends gave him an unfair edge on the public market. That framing sounds familiar from old-school insider-trading cases, but the underlying mechanics are new. The bet was on an event contract, a contract that pays out based on whether a real-world event happens, here the identity of the year's top-trending artist, rather than a stock, and the venue was a crypto-native prediction market rather than the New York Stock Exchange. Federal insider-trading law, built around securities fraud under SEC Rule 10b-5 and the misappropriation of confidential corporate information, has no clean fit for that hybrid, and the Justice Department is now arguing one into existence (NPR; TechCrunch).
The Spagnuolo indictment is not an isolated story. It is the most visible endpoint of a pattern that has been building in public for at least a year. A New York Times investigation cited by CNET in May 2026 flagged more than 11,000 Polymarket accounts whose trading patterns were consistent with insider-style activity, suggesting the engineer case is one node of a much larger network rather than a one-off (CNET). A separate set of high-profile incidents has surfaced the same vulnerability in different guises. U.S. Army special forces soldier Gannon Ken Van Dyke was charged in April 2026 with using classified information about a planned Maduro removal operation to win more than $400,000 on Polymarket — a bet placed, according to prosecutors, after he helped plan the operation and signed nondisclosure agreements about it (CBS News). Former Congressman George Santos was investigated for placing Kalshi bets that he would not attend Trump's State of the Union address, then publicly posted on X that he would be there — a deception that turned profitable when his absence sent odds on his attendance plummeting (NPR). And a Wall Street Journal investigation found evidence that Polymarket ran a secret marketing campaign paying social media influencers to simulate organic trading enthusiasm for products they had already placed (CNET; WSJ via CNET).
What is striking is that none of these incidents required anyone to hack a system or break a protocol. They exploited properties that are designed into the platform. Prediction markets price information, and the person closest to non-public information has the largest edge, so anyone with privileged access to corporate, government, or operational data has a built-in profit channel the moment that data becomes tradeable on a contract. Identity on most prediction markets is pseudonymous at best. Opening an account and funding it does not require the kind of "know your customer" verification that brokerages and banks have run for decades, which means the edge can be monetized without revealing who is taking the other side of the bet. Surveillance on these venues is thin, in part because the contracts are not securities and do not fall under the same reporting regimes that produced the stock-market tools for spotting suspicious patterns (CNET; NPR).
That design is not an accident. Prediction markets work by aggregating dispersed information into prices that look like probabilities, and the platforms argue that any friction on identity or trade size would degrade signal quality. The trade-off is real, but it is also a choice, and it is now producing predictable consequences. Kalshi, one of the two largest U.S.-accessible prediction markets, operates under CFTC oversight as a designated contract market, which gives it some of the surveillance and disclosure obligations that Polymarket, operating offshore with a U.S.-facing front-end, lacks. The two dominant venues are therefore not equivalent. One is regulated like a futures exchange, the other like an offshore crypto casino, and users cannot tell from the homepage which world they are entering (CNET).
The big-tech entry signal sharpens rather than softens the picture. Meta is reportedly building a standalone prediction-market app, internally referenced as Arena, designed to compete with Kalshi and Polymarket and to leverage Meta's social graph as both distribution and identity layer (NPR; Livemint; Memeburn). If Meta ships Arena, the category will gain a billion-user-scale identity layer and an audience accustomed to social-feed-driven financial products, and the regulator-versus-platform contest over what a prediction market is allowed to be will move into a higher gear. Reporting on Arena remains based on internal documents and is not a confirmed launch.
The legal theory the Spagnuolo prosecution advances is the same theory that will decide whether the next engineer, soldier, or congressional insider pays a fine or walks. Insider-trading law in the United States has historically required either a securities-fraud hook under SEC Rule 10b-5 or a fiduciary-misappropriation theory against someone who owed a duty to the source of the information. Event contracts are not securities, so the SEC hook is contested, and the duty owed by a Google engineer to Google's most-searched list is not the kind of duty the doctrine was built around. The Justice Department is essentially asking a court to extend the misappropriation theory into a new asset class, and the outcome of Spagnuolo's case will set the boundary (DOJ SDNY press release; DOJ indictment filing).
For now, three things are worth watching. First, whether the Spagnuolo prosecution survives a motion to dismiss, because a successful 10b-5 or misappropriation theory applied to event contracts would force Kalshi and Polymarket to redesign identity and surveillance in the same direction as equity markets. Second, whether the Commodity Futures Trading Commission, which regulates Kalshi, treats the cross-platform pattern as a compliance problem rather than a competitor's problem and uses its authority to push reporting standards onto the unregulated venues. Third, whether Meta's Arena product, if it ships, comes with identity verification and disclosure obligations baked in or borrows the offshore-platform template.
The user takeaway is concrete. Before putting money on a prediction market, ask whether the venue identifies its users the way a brokerage does, whether it discloses who is taking the other side of large trades, and whether it shares suspicious-activity reports with regulators. Before assuming that a prediction-market price reflects crowd wisdom, ask whether the participants with the largest stakes had access to information the crowd did not. And before treating the next colorful insider-trading headline as a story about one bad actor, ask whether the platform's design made that trade possible in the first place.