Recently, I noticed an interesting incident in the Polymarket prediction market that caught quite a bit of attention. Six accounts are suspected of insider trading and made around $1.2 million by predicting the U.S. attack on Iran at the end of February. The strange thing is that these accounts were created only in that month and only made one trading activity before disappearing.

According to analysis from Bubblemaps, most of these wallets were funded within 24 hours before the attack actually happened. They bought "Yes" shares just a few hours before reports of explosions in Tehran spread. One account even purchased over 560,000 shares at about 10.8 cents per share, paying nearly $560,000 when the market settled at $1. Another account bought nearly 150,000 shares at 20 cents, yielding six-figure profits.

The attack also impacted the broader crypto market. Bitcoin prices dropped while oil futures on Hyperliquid rose significantly. Trading volume on the February 28 contracts reached nearly $90 million, part of over $529 million wagered on the market since December.

Even more interesting, this happened amid increasing regulatory attention to insider trading in prediction markets. The week prior, Kalshi, a competing platform registered with the CFTC, halted and fined two users for insider trading, including a visual editor from MrBeast suspected of trading based on knowledge of event outcomes. Kalshi said they have investigated around 200 cases and have more than a dozen active investigations. Even in tightly regulated areas like the New York ZIP code, a financial hub, concern over illegal trading activities is growing.

The CFTC itself has issued warnings that insider trading on event contracts may violate U.S. laws. They describe exchanges as the "first line of defense" in preventing such activities. In another case, a prospective politician was even sanctioned for betting on their own political race.

The funniest part is that there is evidence that Polymarket traders even engaged in insider trading on markets designed to catch insider trading itself. When blockchain investigator ZachXBT leaked that he would publish findings on the crypto platform Axiom, some people clearly already knew the answer. Lookonchain identified 12 wallets aggressively betting on Axiom before the official announcement was made.

All of this shows that as blockchain adoption grows, the metadata available to machine learning models also advances. Privacy approaches based on obfuscation are structurally declining in effectiveness. Clearly, there are significant regulatory challenges in maintaining the integrity of prediction markets in this digital era.