The surveillance state is expanding, and this time it is arriving not through a government mandate but through a market regulator’s quiet embrace of machine learning. In a move that signals how seriously financial authorities have begun to take the integrity of prediction markets, the US Commodity Futures Trading Commission has begun deploying artificial intelligence tools to monitor trading activity on platforms like Kalshi and Polymarket — a development that carries implications reaching far beyond the borders of Wall Street or Washington.
Prediction markets, for those who have watched them graduate from academic curiosity to financial mainstream, are exchanges where participants trade contracts tied to the outcomes of real-world events: election results, central bank decisions, geopolitical developments, even sporting outcomes. Their proponents argue they aggregate information more efficiently than polling or conventional forecasting, surfacing the collective wisdom of crowds with skin in the game. Their critics have long worried that anyone with genuine non-public knowledge of an upcoming event — a politician’s campaign manager, an aide to a central banker, a company insider — could exploit that information for profit in ways that are difficult to detect and harder to prosecute.
That concern is no longer theoretical. A senior compliance consultant who advises several prediction market operators, speaking on condition of anonymity because of ongoing regulatory discussions, described a case encountered in the past year in which a cluster of accounts began building large positions in a contract tied to a legislative vote roughly forty-eight hours before its outcome was known publicly. “The positions were sized and timed in a way that looked, frankly, like someone knew something,” the consultant said. “The challenge was proving it. These platforms generate enormous amounts of data, and traditional surveillance methods simply cannot process it fast enough to be useful.”
That is precisely the gap that AI-driven surveillance is designed to close. The CFTC’s approach, as described in materials released to industry stakeholders, involves pattern-recognition systems trained on historical trading data that can flag anomalous activity — unusually large position accumulations ahead of known information events, accounts with correlated trading behaviour suggesting coordination, or velocity spikes that precede market-moving news by statistically improbable margins. The models, according to regulatory officials quoted in industry briefings, are designed not to render verdicts but to triage: surfacing cases for human investigators who can then apply legal judgment and pursue formal inquiries.
The timing is not accidental. Prediction markets have grown at a remarkable pace since their legal status in the United States was clarified through a series of CFTC rulings that allowed event contracts on regulated exchanges. Polymarket, which operates under a different regulatory structure that has generated its own controversies, nonetheless saw volumes surpass several billion dollars during the most recent US election cycle. Kalshi, fully regulated under CFTC oversight, has expanded its contract offerings substantially and is seeking to list contracts on a wider range of political and economic events. The combination of scale, political sensitivity, and information asymmetry has created a regulatory environment in which doing nothing is no longer viable.
For Gulf-region financial centres watching this evolution, the implications deserve careful attention. The Dubai International Financial Centre and the Abu Dhabi Global Market have both expressed interest in prediction markets and related instruments as part of broader efforts to position themselves at the frontier of financial innovation. The CFTC’s move toward AI surveillance establishes a template — and arguably a compliance floor — that any jurisdiction hoping to host these markets credibly will eventually need to meet. Regulators who get ahead of the curve on surveillance infrastructure may find themselves better positioned to attract the institutional players who will ultimately determine whether prediction markets become a mainstream asset class or remain a niche curiosity.
There is also a deeper philosophical point embedded in this development. The CFTC’s deployment of AI reflects a recognition that the information advantage problems in prediction markets are not merely similar to those in securities markets — they may in some respects be more acute. A stock price incorporates information about a company continuously over time; a prediction market contract converges on a binary outcome at a fixed future moment, creating sharp incentives to exploit non-public information in the days immediately preceding that moment. The surveillance challenge is therefore both more concentrated in time and more dependent on contextual knowledge that sits outside the market data itself.
Whether AI tools can truly bridge that gap remains an open question. The best pattern-recognition systems in financial surveillance are effective at identifying what looks statistically unusual; they are considerably less effective at distinguishing between illegal information asymmetry and superior analytical skill — which is, after all, what prediction markets are supposed to reward. Getting that distinction right will require not just better algorithms but clearer legal definitions of what constitutes insider information in the context of political and geopolitical events, a question regulators have so far been reluctant to answer directly.
The experiment is nevertheless underway, and the direction of travel seems unlikely to reverse. Prediction markets are growing too large, too visible, and too politically consequential to be left outside the perimeter of active regulatory oversight. The question is no longer whether they will be watched — it is how well the watchers can keep up.