Hook
Contrary to popular belief, prediction markets are not immune to insider trading.
They are just vulnerable in a different dimension.
On February 18, 2025, ABC News reported that Tyler Perez, a 39-year-old teleprompter operator for the White House, made over $100,000 trading on Kalshi—a CFTC-regulated prediction market—by betting on specific words that Donald Trump would utter during his speeches.
Perez had access. Real-time access. The teleprompter script.
He placed bets on markets like "Will Trump say 'infrastructure' tonight?" Then, mid-speech, as the president read the line, Perez closed his positions at maximum profit. The trades were textbook insider transactions: non-public information, clear intent, immediate execution.
Kalshi’s monitoring team flagged them. They reported to the CFTC. Perez is now in settlement negotiations.
But the story is not about a rogue operator. It is about a structural flaw in the architecture of event contracts—a flaw that no audit report has ever flagged.
Context
Kalshi is a designated contract market (DCM) regulated by the Commodity Futures Trading Commission. It allows users to trade binary event contracts: yes/no bets on the outcome of political events, economic reports, and even presidential speeches. Unlike Polymarket, which runs on-chain with smart contracts and pseudonymity, Kalshi operates with full KYC/AML, real-name registration, and a centralized order book.
Since its launch, Kalshi has marketed itself as the "safe" prediction market. It submits to US regulation. It reports suspicious activity. It even requires users to disclose their employers—a rule added just last month, after this incident began to surface.
The market that Perez exploited was a "Mentions" market: traders speculate on whether a specific word will be spoken during a live event. The contract resolves when the event ends. The payout is binary. If you know the answer before the audience hears it, you have an informational advantage roughly equivalent to insider knowledge in a stock.
Perez placed multiple bets on multiple speeches, including the 2024 State of the Union. His peak profit: over $100,000. But he made one mistake: he closed his positions while the speech was still airing. That real-time action—the withdrawal of a bet while the event is in progress—is a red flag. A trader who knows the outcome in advance will not hold until resolution; he will exit early to lock in profits, especially if the odds shift as the event unfolds.
Kalshi’s monitoring team caught this pattern. They flagged the accounts. They alerted the CFTC. Perez was placed on administrative leave by the White House.
This is the first major insider trading case in a US-regulated prediction market. But it is not the first prediction market insider trading case overall. The FBI has investigated two others: one involving a Venezuelan official betting on Maduro’s political decisions, and another involving a Google employee trading on internal product leaks. The pattern is consistent: information asymmetry, real-time exploitation, and a platform that only catches the crime after the fact.
Core: The Bytecode of the Bet
Let me take you behind the code, because that is where the real risk lives.
Kalshi does not publish its smart contract source code. It is a centralized platform. But the logic of a "Mentions" market can be formalized as a state machine:
