Insider Trading on Prediction Markets: The 2026 Reckoning

Last updated: August 2026  ·  10 min read

In April 2026, a federal indictment alleged that a US Army master sergeant had used classified information about a military operation to place prediction market positions worth more than $409,000 in profit. A month later, the House Oversight Committee opened a formal investigation into Polymarket and Kalshi. By August, the two platforms had between them flagged more than a hundred potential insider trading cases in a single year.

This looks like a scandal, and in the narrow legal sense it is. But it is also something more interesting: a collision between what prediction markets are designed to do and what society is willing to let them do. Understanding that tension matters more than tallying the cases, because it determines what the next generation of forecasting platforms will have to look like.

What Actually Happened in 2026

The year produced a cluster of cases that were difficult to dismiss individually and impossible to dismiss collectively.

On 22 May 2026, House Oversight Chairman James Comer opened an investigation, requesting information from both platforms on identity verification for domestic and international account holders, enforcement of geographic restrictions, and methods for detecting anomalous trading patterns. Committee representatives indicated subpoenas remained possible.

Pressure has continued to build from other directions. In August 2026 the New York City Council opened its own inquiry into how prediction market platforms advertise within the city, signalling that scrutiny is broadening from trading conduct into marketing practice.

The Uncomfortable Part of the Theory

Here is where the story becomes genuinely difficult rather than merely scandalous.

Prediction markets are valuable precisely because they draw private information into a public price. Someone who knows something the crowd does not is supposed to trade on it, and in doing so move the probability toward the truth faster than surveys or commentary could. Robin Hanson, the economist most associated with the theoretical foundations of these markets, has argued that informed trading is not a defect in the mechanism — it is the mechanism.

By that logic, a market that successfully excluded all privately informed participants would be a market that had removed its own reason to exist. It would aggregate nothing but public opinion, which is a poorer forecast than the one it replaced. This is not a fringe position; it follows directly from the information-aggregation argument that justifies prediction markets in the first place, which we examine in how accurate prediction markets actually are.

Where the Theory Stops Working

The Hanson argument holds for a particular kind of information: privately gathered insight, superior analysis, specialised expertise, or local knowledge. A logistics analyst who correctly reads shipping data is exactly the participant the mechanism wants.

It stops working in at least three situations, and the 2026 cases fall almost entirely into them.

When the information is classified

Trading on knowledge of an impending military operation does not improve public forecasting in any meaningful sense. The information cannot be legally acted upon by anyone else, so the price movement conveys a signal the public cannot interpret or verify. It also creates a direct financial incentive to leak state secrets, which is a harm entirely separate from market quality.

When the trader controls the outcome

A candidate holding a position on their own race is not forecasting — they are influencing the very event the contract resolves against. This breaks the assumption that participants observe outcomes rather than determine them. The same objection applies to any official positioned to affect a decision the market is pricing.

When the information arrives through a duty of confidence

A staff member who learns something because of their role, under an implicit or explicit obligation not to disclose it, is not contributing independent insight. They are converting institutional trust into private gain — the conventional objection to insider trading in securities, and it transfers cleanly.

The Enforcement Gap

The numbers reveal a structural problem. Platforms identified over a hundred potential cases in 2026. The CFTC brought civil action against three traders. That is not primarily a failure of will — it reflects an agency with finite capacity supervising a market that has grown extremely fast, while simultaneously contesting jurisdiction with individual states over whether these products are lawful at all.

The result is an awkward equilibrium in which detection has improved considerably faster than consequence. Platforms increasingly find the suspicious activity; comparatively little follows. That gap is itself a risk, because it invites the assumption that the rules are unenforceable. The broader regulatory picture, including the ongoing federal-versus-state dispute, is covered in our guide to prediction market legality in 2026.

What This Implies for Market Design

If the problem is partly structural, then surveillance alone cannot solve it. Detection after the fact is necessary but reactive. The more durable lever is design — deciding in advance which questions a platform lists and how they resolve.

These are not exotic requirements. They are the same considerations that separate a well-constructed market from a poorly constructed one in ordinary conditions, examined in more depth in what makes a good prediction market. The 2026 cases simply raised the cost of getting them wrong. Platforms like Nexory that focus on outcomes shaped by broad public information rather than by a handful of officials face a structurally narrower version of this problem — though no platform listing real-world events is fully insulated from it.

Conclusion

The 2026 insider trading cases are not evidence that prediction markets do not work. In a sense they are evidence that the mechanism works exactly as designed — private information moved prices, and it moved them early. The problem is that some categories of private information should never have been tradeable in the first place.

The likely outcome is not the end of prediction markets but a narrowing of what they are permitted to price, and a raising of the standard for how those questions are constructed. Platforms that anticipate that shift will be better positioned than those waiting to be told. What remains genuinely uncertain is where regulators eventually draw the line, and whether enforcement capacity ever catches up with detection.