Prediction markets once promised a cleaner signal than polls or pundits. Now they face accusations of becoming conduits for inside information. Traders on platforms such as Polymarket have scored outsized wins on events ranging from military operations to corporate product launches. Regulators notice. So does Wall Street.
The Commodity Futures Trading Commission has begun feeding vast streams of trade data into artificial intelligence systems. The goal is straightforward. Spot patterns that suggest traders acted on nonpublic knowledge. CFTC Chairman Michael Selig told WIRED that the volume overwhelms traditional review. “You’ve got so much data. When we feed it into AI, we get really great information. It can help us understand things, like where we might want to investigate, or when we might need to send a subpoena to a trader.”
Such comments mark a shift. For months, stories of suspiciously timed bets multiplied. One US Army special forces soldier faced arrest in April after allegedly using classified details about a Venezuela operation to win more than $400,000 on Polymarket contracts. Prosecutors said the bets preceded public announcements. The platform later claimed it had flagged the activity.
But that case stands almost alone in court records. Selig informed Congress the agency pursues “hundreds, if not thousands” of tips. Many involve offshore venues that US users reach through virtual private networks. Polymarket’s crypto-based exchange remains blocked for American traders. Enforcement still requires connecting wallets to real identities. Here the AI enters.
The CFTC combines its own surveillance software with tools from Chainalysis and Nasdaq Smarts. Blockchain analytics help trace wallet clusters that appear suddenly, bet heavily, and vanish after an event resolves. Chainalysis spokesperson Maddie Kenney explained that the firm organizes data for both the agency and the platforms themselves. “The value Chainalysis adds for our customers, including Polymarket and the CFTC, is organizing the data and enriching it with the attributions and insights we’ve accumulated over years in the space.”
Platforms have responded. In April Polymarket announced its own deal with Chainalysis to monitor for manipulation and update market integrity rules. Competitor Kalshi, which operates under US regulation, publicized suspensions and penalties against users flagged for insider activity. Neither company answered every question from reporters. Yet both now advertise compliance efforts that would have seemed unnecessary a year ago.
And the volume keeps rising. Goldman Sachs chief US equity strategist Ben Snider described to InvestmentNews how his team processes signals from multiple markets. “I find myself relying on them more and more as an indicator of what the markets are really thinking. I have AI combine Kalshi, Polymarket, anything else I can find, and aggregate them across my screen.”
Private firms face parallel pressure. OpenAI fired an employee last winter after an internal probe found the person used confidential details to trade on prediction contracts. Unusual Whales, a financial data service, identified clusters of wallets that bet on OpenAI product releases hours before announcements. One analysis counted 60 suspicious wallets tied to unreleased projects since 2023.
Similar patterns surfaced around Google’s Year in Search rankings. A trader known as AlphaRaccoon reportedly turned roughly $10,000 into nearly $200,000 on a single correct outcome involving an obscure musician. The same account nailed 22 of 23 specific search rankings. Google had accidentally leaked elements of the list. Questions remain whether the trader simply guessed well or enjoyed an early look.
Senator Chris Murphy of Connecticut raised alarms in March. He told reporters he suspected White House staff might have traded on war-related events. Weeks later seven lawmakers sent a letter urging the CFTC to examine overseas contracts tied to military action. They called some of the bets “morally obscene.”
Selig’s agency claims broad reach. The 2010 Dodd-Frank Act expanded its power over foreign swaps that affect US markets. “We’re surveilling the markets on a global basis,” he said. Yet he also described limits. Extraterritorial cases proceed only when victory seems likely. Otherwise the CFTC refers matters to foreign counterparts. “In any extraterritorial litigation, there’s going to be challenges to our authority, and that could also impair our ability to bring cases in the future.”
The agency itself remains short-handed. Selig acknowledged it is “especially lean right now” but is hiring. AI fills gaps. It flags anomalies faster than humans can scan millions of trades. Still, false positives worry compliance officers. Legitimate sharp traders exist. Distinguishing them from insiders requires context that algorithms struggle to supply without human review.
Consultants already advise banks on the new risk. Carlo di Florio, president of ACA Group, described SEC questions that now land on trading desks. “The SEC is very focused on how are you as a firm getting comfortable that your employees aren’t walking in the morning and taking sensitive firm information and placing bets on Polymarket and Kalshi. What are you doing to prevent that?”
Prediction markets trade on everything from election outcomes to weather readings to corporate earnings. Speed matters. A well-timed bet can pay out within hours. That velocity, combined with pseudonymity on crypto platforms, creates openings. Bots have entered the field too. Some reports suggest automated strategies now dominate certain contracts, amplifying volatility when news breaks.
Yet the information generated still holds value. Investors scan aggregated probabilities for hints about Federal Reserve moves or supply chain disruptions. Snider cited one recent example involving the Strait of Hormuz. Prediction prices suggested reopening timelines that shaped his team’s oil forecasts. The AI layer simply makes the data digestible.
Critics counter that tolerating some insider flow corrupts the entire mechanism. If a few participants possess material nonpublic information, prices reflect privilege rather than collective wisdom. Platforms once argued that such bets improved accuracy. Most have dropped that defense. Polymarket’s CEO Shayne Coplan once spoke favorably of insider participation. The company’s recent partnerships signal a different stance.
Legal scholars watch how courts treat these cases. The soldier charged in the Venezuela matter represents the first clear criminal insider trading prosecution tied to a prediction platform. More will follow if the CFTC’s AI pipeline produces actionable leads. Subpoenas to exchanges and wallet providers have already increased.
Technology cuts both ways. Traders deploy their own models to scan for inefficiencies or front-run news. Regulators answer with more sophisticated detection. The result looks like an arms race played out in code. One side hunts edges. The other hunts the hunters.
For now the CFTC projects confidence. Selig promised action against offenders “no matter how large or how small.” Whether AI delivers enough precision to support that claim will determine if prediction markets mature into respected forecasting tools or remain shadowed by doubt. The data keeps pouring in. The machines keep watching.
WebProNews is an iEntry Publication