Prediction Markets Have an Insider-Information Problem

Strategic Analysis | September 14, 2026

Prediction markets are supposed to reward people who know something the rest of the market does not.

That is also becoming their biggest problem.

New blockchain analysis reportedly identified 19 interconnected Polymarket accounts that collectively went 41-for-42 on earnings bets involving 18 companies audited by KPMG, generating about $22,000 in profit. The companies included Wells Fargo, Home Depot, DoorDash, and General Mills. The trades began in November 2025 and continued for several months.

That is an extraordinary batting average.

It is not, by itself, proof of insider trading.

Blockchain-forensics firm Bubblemaps identified links among the accounts, but neither that analysis nor the win rate establishes who controlled them, how the traders obtained their information, or whether anyone breached a legal duty. No KPMG employee has been charged in connection with the cluster, and it has not been established that the accounts belong to the KPMG employee whom federal authorities have separately been investigating over prediction-market trading. KPMG has said it has zero tolerance for employees using nonpublic client information to trade.

But the pattern matters even if no unlawful trading is ever proved.

Because the innovation of prediction markets is not merely that they put a price on the future.

They put a price on secrets.

And once almost any future event can become tradable, almost any piece of confidential information can become financially valuable.

That creates a market-integrity problem much larger than KPMG.

Ordinary financial markets already live with information asymmetry.

One investor may understand a company better than another. A hedge fund may purchase alternative data, commission proprietary research, interview suppliers, build a better model, or simply interpret public information more intelligently.

That is not a defect. It is part of what makes a market work.

Prediction markets take that principle further.

A trader who knows more about an election, a football team, an earnings report, a regulatory decision, an award show, or an impending corporate announcement can move the market toward a more accurate price.

Polymarket describes its prices as expressions of the market’s collective probability assessment. A contract trading at 65 cents effectively implies that the market assigns roughly a 65% probability to that outcome.

The more informed the traders, the better that forecasting mechanism should become.

But there are two very different ways to know more than everyone else.

You can know more because you did better work.

Or you can know more because someone entrusted you with the answer.

Those two information advantages may look identical on a trading screen.

Legally and institutionally, they are worlds apart.

That distinction may become the central integrity question for prediction markets.

The phrase insider trading can create the wrong impression.

American law does not generally require every trader to possess the same information.

The important question is often how the trader obtained the information and what duties came with it.

The Commodity Futures Trading Commission has now taken that principle directly into prediction markets.

In February, the CFTC’s Enforcement Division issued an advisory explaining that misuse of confidential information in breach of a pre-existing duty of trust or confidence can violate Section 6(c)(1) of the Commodity Exchange Act and CFTC Rule 180.1—the commodities-law analogue to the misappropriation theory familiar from securities law.

The concept is relatively intuitive.

If an employee receives confidential information because of a job, owes a duty to protect it, and instead uses it to trade for personal profit, the problem is not merely that the employee knew more.

The problem is that the employee converted someone else’s confidential information into a personal trading advantage.

The CFTC has used the same theory outside prediction markets. Earlier this year, a federal court entered a consent order in CFTC v. Clark involving the misappropriation of confidential commodities-trading information in breach of a duty to an employer.

Prediction markets simply create many more places where that theory might matter.

And federal authorities have already begun testing it.

In April, prosecutors charged Army Master Sergeant Gannon Ken Van Dyke with offenses including commodities fraud after alleging that he used classified knowledge concerning a U.S. military operation in Venezuela to purchase Polymarket contracts involving U.S. forces and Nicolás Maduro. According to the government, approximately $33,000 of trades produced roughly $410,000 in profits. The allegations remain contested.

A month later, federal prosecutors charged Google engineer Michele Spagnuolo with commodities fraud, wire fraud, and money laundering, alleging that he used confidential internal Google data to make approximately $1.2 million on Polymarket contracts involving Google's Year in Search results. Spagnuolo has challenged the government’s theory.

Those cases are important for another reason: the legal foundation is not completely settled.

The government’s position is that these event contracts fall within the Commodity Exchange Act and that its antifraud provisions apply. The CFTC has expressly said that prediction-market event contracts can be swaps and that Rule 180.1 reaches insider trading in them.

Defendants are challenging that premise, arguing in part that certain Polymarket wagers are not “swaps” within the meaning of federal commodities law at all.

So there are really two fights developing simultaneously.

One asks:

What conduct counts as unlawful misuse of confidential information?

The other asks an even more foundational question:

Does the federal commodities regime reach the particular prediction-market contract in the first place?

Courts have not finished answering the second question.

The emerging market rules may be just as important as the litigation.

Polymarket now expressly prohibits trading on confidential information when using it would breach a pre-existing duty of trust or confidence. It also prohibits trading on improperly tipped confidential information and trading by someone with sufficient authority or influence to affect the event’s outcome.

But Polymarket does not prohibit every private informational advantage.

Its published examples specifically allow a trader to act on proprietary analysis derived from public data, a privately commissioned poll, personally developed sports statistics, or even information lawfully overheard without violating a confidentiality duty.

That distinction is crucial.

A prediction market that prohibited anyone from trading unless everyone knew the same thing would defeat much of the purpose of having the market.

The valuable line is therefore unlikely to become:

Public information good; private information bad.

It is more likely to become:

Legitimately acquired informational advantage good; misappropriated or outcome-controlled information bad.

The CFTC itself has acknowledged the tension. In its current prediction-market rulemaking, the agency asked whether there can be public value when traders with asymmetric information participate because their trades may make market probabilities more accurate—while simultaneously recognizing that those advantages can create unfairness, manipulation, and misuse of inside information.

That is not a minor regulatory detail.

It is the paradox at the center of the product.

The market becomes more informative when people who know more trade. The market becomes less trustworthy when people are being paid for secrets they were supposed to keep.

The KPMG-linked pattern matters because it demonstrates what market surveillance increasingly has to detect.

The reported accounts did not simply win often. Bubblemaps reportedly identified 19 connected accounts, and the 42 earnings wagers all involved companies audited by the same accounting firm. Researchers had reportedly begun flagging the pattern months before the latest reporting.

Again, none of that proves wrongdoing.

There are innocent explanations for unusual trading success. A skilled analyst can outperform. Multiple wallets can have legitimate connections. Small sample sizes can produce remarkable streaks. A trader can legally develop proprietary information advantages.

But sophisticated surveillance does not wait for a confession.

It looks for combinations of signals: unusually high win rates, common funding sources, wallet relationships, concentration around one information nexus, unusual trading immediately before announcements, repeated success in low-probability outcomes, and behavior inconsistent with the trader’s performance elsewhere.

The blockchain creates an unusual advantage here.

Polymarket’s international platform settles trades on-chain, meaning a substantial amount of transactional activity can be publicly examined. Polymarket itself emphasizes that transparency and says it supplements blockchain visibility with real-time surveillance and outside specialists.

But transparency is not the same thing as identification.

A blockchain can show that Wallet A sent money to Wallet B.

It does not necessarily tell you who was sitting behind the keyboard.

That means the next generation of prediction-market surveillance cannot merely be good blockchain analytics.

It increasingly requires what mature financial markets already rely on: identity data, relationship mapping, employer information, restricted-person lists, account-link analysis, cross-market surveillance, suspicious-activity escalation, cooperation with event sources, and mechanisms for preserving evidence before it disappears into another wallet.

Prediction markets may have been born from crypto.

To become serious financial infrastructure, their compliance systems may need to look increasingly like Wall Street.

This is where the KPMG story becomes a sports story.

The sports-betting industry has dealt with inside information for years. But traditional sportsbooks largely monetize information about athletic competition.

Prediction markets can monetize almost anything surrounding it.

An undisclosed injury.

A trade that has been agreed to but not announced.

A disciplinary ruling sitting on a league executive’s desk.

A coach who has privately accepted another job.

A player who has decided to enter the transfer portal tomorrow.

An NCAA eligibility decision awaiting publication.

A league expansion vote.

A draft decision.

A media-rights announcement.

A collective-bargaining development.

A stadium decision.

The people who know those answers first are not limited to athletes.

They may include trainers, physicians, agents, lawyers, equipment staff, league employees, athletic-department administrators, compliance personnel, broadcasters, consultants, investigators, technology vendors, family members, and countless others who never previously had a meaningful way to monetize that information.

Prediction markets change the economics of confidentiality.

A trainer knowing that the star quarterback will not play on Sunday has always possessed valuable information.

But if a market can be created not only on the game but directly on whether the player will appear, whether he will be listed as inactive, whether the team will announce an injury by Friday, or whether a roster move will occur, that information becomes much easier to convert directly into cash.

The CFTC has already identified precisely this risk. In remarks this spring, an agency official used sports injuries as an example, noting that a trainer with advance nonpublic knowledge could potentially trade on that information in violation of a confidentiality duty.

The sports industry is responding.

The CFTC has entered information-sharing arrangements with Major League Baseball and the NHL aimed at protecting both competition and prediction-market integrity.

The NCAA has gone further, urging the CFTC to require prediction markets to monitor and promptly report suspicious activity—including anomalous trading suggesting use of material nonpublic information—and initially asked regulators to suspend collegiate sports contracts until stronger safeguards exist.

That is where this is heading.

Sports leagues once needed integrity systems capable of detecting a player betting against his own team.

They may soon need systems capable of determining whether a graduate assistant’s roommate traded on advance knowledge of an eligibility ruling.

That is a radically larger surveillance problem.

There is a serious counterargument.

If prediction markets exist to aggregate dispersed information, excluding informed participants can make the product worse.

Someone who has studied every public injury report, developed a proprietary model, purchased better data, commissioned independent research, or noticed something everyone else missed should ordinarily be allowed to capitalize on that insight.

Markets work because people have different information.

The goal cannot be to make everyone equally ignorant.

There is also a danger in borrowing securities-law concepts mechanically. A public company has officers, directors, shareholders, mandatory disclosure systems, and an established body of securities regulation. A football injury, presidential statement, Oscar result, papal election, weather event, or coaching decision may have none of those things.

Who exactly is the “insider”?

What information belongs to whom?

When does an informal expectation of secrecy become a legal duty?

What if someone learns the information through observation rather than disclosure?

What if the trader helps cause the event?

Prediction markets turn questions that securities law spent decades refining into questions that can arise in nearly every corner of ordinary life.

That is precisely why a blanket insider prohibition will not solve the problem.

The more durable framework is likely to focus on three categories: misappropriated confidential information, improperly tipped information, and trading by people who can control or materially influence the outcome.

Everything outside those categories becomes harder.

And that uncertainty is not a bug that regulators can simply write away. It is inherent in markets whose underlying “assets” are events rather than companies.

Prediction-market growth should already be changing corporate and sports compliance.

An employee policy that says “do not trade our client’s stock” may no longer be enough if the employee can instead buy a contract on whether the client will beat earnings.

A law firm’s securities-trading restrictions may need to consider contracts tied to litigation outcomes, deals, government approvals, or client announcements.

A professional-services firm may need to address event contracts involving clients whose confidential information passes through the firm.

Sports organizations may need policies broad enough to cover prediction contracts involving injuries, transactions, discipline, employment decisions, eligibility, and league operations—not merely wagers on games.

And exchanges will face pressure to identify categories of contracts where informational asymmetry is unusually dangerous.

Not every market presents the same risk.

Weather forecasting may reward private models without creating meaningful confidentiality problems.

A contract on tomorrow’s confidential corporate announcement is different.

The closer the underlying event is to information controlled by an identifiable person or small group, the stronger the case becomes for enhanced surveillance, restricted-person rules, reporting relationships, lower position limits, delayed listings, or in some cases not listing the contract at all.

Prediction markets like to emphasize that virtually anything can become a market.

That may be technologically true.

It does not follow that everything should be governed by the same integrity architecture.

The immediate legal battlefield will be the enforcement cases already underway.

The Van Dyke and Spagnuolo matters could help determine whether federal courts accept the government’s theory that the relevant Polymarket contracts fall within the Commodity Exchange Act. If courts reject that premise, regulators may need new statutory authority—or Congress may eventually have to decide expressly how offshore and domestic event contracts fit within federal law.

At the same time, the regulated U.S. market continues expanding. Polymarket US is now a CFTC-designated contract market, while multiple exchanges are listing sports event contracts and mainstream financial platforms are making prediction markets easier for retail customers to reach.

That combination—greater liquidity, easier access, more categories of events, and more money—makes sophisticated surveillance unavoidable.

The KPMG-linked accounts generated only about $22,000.

That is almost beside the point.

The important number is 41 out of 42.

Not because it proves somebody cheated.

Because it demonstrates what these markets look like when a trader appears to know the answer before everyone else.

Sometimes that will be brilliant analysis.

Sometimes it will be luck.

Sometimes it may be information legitimately acquired through hard work.

And sometimes it may be somebody else’s secret.

The future of prediction markets may depend on whether exchanges and regulators can reliably tell the difference.

Prediction markets promise to turn dispersed information into useful prices.

That promise only works if people trust the process producing those prices.

The industry therefore faces a tension more fundamental than the familiar debate over whether prediction markets are gambling or derivatives.

A prediction market’s value comes from information asymmetry. Its integrity risk comes from information asymmetry too.

The answer cannot be to eliminate informational advantage. That would eliminate much of the product’s usefulness.

The challenge is building a system that rewards people for discovering the future without rewarding them for stealing it.

And as prediction markets expand from elections and football games into earnings reports, military operations, injuries, corporate announcements, eligibility rulings, and everything else that can be phrased as a yes-or-no question, that distinction will become harder—and more important—to police.

Because once every future event has a price, every secret has one too.