Prediction Markets Log Trades, but Insiders Stay Hidden

Insider Information Trouble Hits Prediction Markets Too

On February 27, 2026, WIRED reported that OpenAI had fired an employee for using non-public company information in prediction market trades. The company explained that using internal information for personal gain violates its internal policy. It didn’t disclose the employee’s name or the specific details of the trades.

Information that only some employees know in advance—like a company’s product launch date—can also become a subject of trading in prediction markets. A user guessing the launch date from news reports and an employee with access to the internal schedule can end up buying and selling the very same contract. I think this information gap is the first issue we need to examine when looking at prediction markets.

Prediction markets are places where people trade contracts on whether a specific event will occur. For example, if a contract that pays $1 when an event happens is trading at $0.7, that price is often read as the market’s implied probability of about 70%. But this is a price that reflects the judgment of the people participating in the trade. It doesn’t always match an objectively verified probability of occurrence.

Personally, when I look at these services, they sometimes feel closer to an “odd-or-even guessing market.” If you actually go in and look, you can bet money on all kinds of topics—not just elections or economic indicators, but product launches, wars, even a celebrity’s content. The broader the range of questions, the broader the pool of people who get to know the relevant information first.

Suspicious trades don’t reveal who’s behind them

Polymarket’s blockchain records let you trace wallet addresses and trading flows. Patterns like new accounts popping up right before an announcement, or a wallet betting big on one event and then going silent, can be leads worth investigating.

An account nicknamed “Google Whale” drew that kind of suspicion. There were reports that it made big profits betting on Google-related outcomes. But a high win rate and strong returns alone can’t confirm that the person was a Google employee, or that they illegally used non-public information.

This distinction matters for other analyses too. The financial data service Unusual Whales flagged 77 positions across 60 wallets trading OpenAI-related events as suspicious cases, based on factors like when the wallets were created, their trading history, and the amounts wagered. A tool flagging something as suspicious and a law enforcement agency proving wrongdoing are two entirely different stages.

A wallet address alone can’t tell you the trader’s real name or employer. You’d need to establish who actually used the account, what information they had access to and when, and whether they were under any obligation to keep that information confidential. That often requires records beyond the trading ledger itself—things like communication logs or account registration details.

We need to separate firings, exchange penalties, and criminal charges

The cases from early 2026 show that responses are already unfolding through multiple channels.

OpenAI’s firing was action the company took under its internal policies. It doesn’t by itself establish criminal guilt. Kalshi, by contrast, investigated a violation of exchange rules and imposed usage restrictions and financial penalties.

Documents released by the U.S. Commodity Futures Trading Commission (CFTC) on February 25 describe a case involving a video editor who traded on a contract related to the YouTube channel he worked for. Kalshi determined there was reasonable basis to believe he had used non-public information he’d learned through his job.

The penalty was a two-year suspension from the exchange plus a total financial burden of $20,397.58 — $15,000 of that was a fine, and $5,397.58 was disgorgement of trading profits. Media reports identified the editor as someone who had worked on content production for MrBeast. This is distinct from any criminal fine that might be handed down by a court.

In Israel, the matter escalated to criminal charges. According to a February 12 Jerusalem Post report, a civilian and a military reservist were indicted for placing bets on Polymarket using classified military information. The report said prosecutors applied charges including serious security breaches, bribery, and obstruction of justice. Much of the case remains under seal — including the defendants’ identities and the specifics of what they bet on.

When information related to military operations becomes a tool for making money, it’s not just market fairness at stake — operational security is on the line too. Still, we need to read the disclosed allegations and what’s actually been proven as separate things.

Having a legal basis doesn’t excuse you from proving and enforcing it

It’s hard to describe America’s prediction markets in exactly the same terms as its stock market. For event contracts on registered exchanges like Kalshi, the Commodity Exchange Act and CFTC regulations serve as the key standards.

In a February 25, 2026 advisory, the CFTC reaffirmed that it has enforcement authority over the use of confidential information in breach of an existing duty of trust or confidentiality owed to the source of that information. It cited Section 6(c)(1) of the Commodity Exchange Act and Regulation 180.1 as the relevant basis. It also noted that exchanges themselves carry obligations around recordkeeping, surveillance, and rule enforcement.

So it’s not accurate to say prediction markets have no legal basis at all for addressing insider trading. What’s needed, case by case, is establishing who breached what duty, and which specific regulations apply to the trade in question. When a platform operates out of another country, or when a trader’s identity is never revealed, the investigation only gets more complicated.

There are precedents from other markets where existing fraud statutes were applied. The U.S. Department of Justice announced in 2023 that Ishan Wahi, a former Coinbase employee who tipped off an acquaintance about upcoming token listings, was sentenced to 2 years in prison on a charge of conspiracy to commit wire fraud. That’s not a prediction-market ruling, but it shows that legal avenues for addressing misuse of confidential information aren’t limited to securities regulation alone.

There’s also been legislative movement aimed at restricting trading by government officials. On January 9, 2026, U.S. Representative Ritchie Torres introduced a bill that would bar public officials and others with access to material nonpublic information obtained through their work from trading related prediction-market contracts. Introducing a bill, of course, is a different thing from it becoming law.

One of the triggers behind this push was suspicious trading ahead of the operation to arrest Maduro. The congressman’s office pointed out that a newly created account placed bets exceeding $30,000 and then received roughly $400,000 in return. There was no announcement at the time confirming whether that account was actually operated by a government official.

Oswarld’s Lens

I have doubts about the way prediction markets frame their appeal to accuracy. Sure, when someone trading has information others don’t, market prices can reflect outcomes faster. But that doesn’t automatically make the way that information was obtained—or the fairness of the trade itself—legitimate.

Polymarket CEO Shayne Coplan, in a CBS interview in November 2025, touted the advantages of an information edge, while also saying that the boundaries of what’s permissible and the ethical standards need to be made explicit. I think we need to actually watch how those standards get applied in practice.

Someone analyzing publicly available data and someone who learned an unreleased result through their job start from very different positions. If ordinary users participate without knowing this gap exists, the risk of them absorbing losses could become a bigger problem than whatever predictive value the market claims to offer.

Eric Zitzewitz of Dartmouth raised a similar concern in comments submitted alongside Rep. Torres’s bill announcement. Insiders’ profits, he noted, can come directly from the counterparties who placed orders at prices set before the news broke—and as concerns about this kind of trading grow, people may simply place fewer orders. Less participation weakens the market’s predictive function itself.

I don’t think technology that transparently records trades solves this problem on its own. You still need to spot suspicious patterns in the records, identify who actually made the trade, and investigate where the information came from and whether any duty was breached.

When evaluating a prediction market, we should look beyond its hit rate to who is trading on what information, and how the platform manages that asymmetry. The CFTC materials below are a good starting point for seeing exactly what conduct regulators flagged as of February 2026.

The draft looks accurate and complete. No corrections needed.

References & Further Reading

• WIRED, OpenAI Fires an Employee for Prediction Market Insider Trading, February 27, 2026. Reports on the company’s confirmation of the firing and analysis of the suspicious trades.
• CFTC, Prediction Markets Advisory, February 25, 2026. Explains two disciplinary cases involving Kalshi and the CFTC’s grounds for enforcement.
• The Jerusalem Post, Israel indicts reservist, civilian for using classified information to bet on IDF military action, February 12, 2026. Covers the indictment tied to classified military information and how much was public at the time.
• Office of Rep. Ritchie Torres, announcement of the Public Integrity in Financial Prediction Markets Act, January 9, 2026. Details the bill’s scope, its rationale, and comments from Eric Zitzewitz.
• U.S. Department of Justice, announcement of sentencing for former Coinbase employee Ishan Wahi, May 9, 2023.
• CBS 60 Minutes, interview with Shayne Coplan, November 30, 2025.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.