A White House teleprompter operator, Pablo Perez, made over $100,000 on Kalshi prediction markets by betting on the exact words his boss would speak. The specific market: “Mentions” contracts, where users wager on whether a politician says a particular term during a speech. Perez, who had access to the teleprompter scripts hours before delivery, placed trades on phrases like “inflation” and “border” ahead of President Trump’s State of the Union address. He even withdrew some bets midway through the speech when he saw the script changing in real time. Kalshi’s compliance team flagged the accounts and reported the activity to the CFTC, which is now negotiating a settlement with Perez. Watching the ledger breathe beneath the noise, this is not just a scandal; it is a stress test for the entire premise that prediction markets can price truth without leaking through the cracks of human access.
To understand the stakes, we must first understand Kalshi’s place in the ecosystem. Launched in 2020, Kalshi is a CFTC-regulated designated contract market (DCM) that offers event contracts on everything from interest rates to political outcomes. Unlike its main competitor Polymarket—which operates as a decentralized, permissionless platform on the Ethereum blockchain—Kalshi enforces KYC/AML and charges fees to cover compliance costs. Its “Mentions” market is a niche but growing product: users bet on whether a specific word appears in a speech, a press conference, or a government report. The appeal is pure information asymmetry—anyone with prior knowledge of the script has an extraordinary edge. Kalshi has rules against trading on non-public information, and as of last month, it now requires users to disclose their employer. Yet Perez still slipped through, because the system relies on self-reporting and post-trade pattern recognition.
Here is where my analysis diverges from the headline. From my years auditing prediction market protocols, I have seen that the greatest risk is not from code exploits but from human behaviour. Kalshi’s detection of Perez came because their monitoring team cross-referenced his employee status with his trading patterns—a process that requires both a culture of surveillance and a willingness to alert regulators. But the “Mentions” market is structurally vulnerable: it converts privileged verbal information into a tradable asset, and no amount of smart contract auditing can prevent a user from reading a script and placing a bet. The core insight is this: prediction markets are not merely markets on future events; they are markets on the distribution of information at a given time. When one participant holds a deterministic advantage over the outcome, the market ceases to function as a price-discovery mechanism and becomes a rent-extraction tool. We minted souls but forgot the container—the container being the ethical framework that separates speculation from exploitation. In my CBDC research, I have observed that the tension between privacy and transparency is the central design challenge of all financial systems. Here, Kalshi chose transparency (employer disclosure) but at the cost of user privacy, and even that was not enough to prevent the trade. The real failure is not technological but procedural: no system can fully prevent a willing insider from using their knowledge, unless the market itself is designed to be unresponsive to isolated advance information. That would require markets that are delayed, or aggregated across many independent sources—neither of which is the case for “Mentions” contracts.
The contrarian angle—the one that cuts against the “prediction markets are broken” narrative—is that this event actually validates the regulated model. Perez was caught, flagged, and faces a CFTC settlement. The same trade on Polymarket would have been anonymous, with no employer disclosure and no monitoring team. The CFTC’s involvement is a double-edged sword: it signals that the government will enforce rules against insider trading, which may deter institutional participants, but it also creates a precedent for collaboration between regulators and platforms. Between the code and the conscience lies the gap—and in that gap, Kalshi chose to close it. Rather than weakening the case for prediction markets, this incident strengthens the argument that compliance-first platforms can coexist with decentralized ones, each serving different risk tolerances. The blind spot, however, is that the market’s design itself enabled this. The “Mentions” market is a textbook case of a market that is fundamentally dependent on information asymmetry—it is not a prediction market about an event, but a market about a predetermined script. Kalshi may need to reconsider whether such markets should exist at all, or whether they require additional safeguards like mandatory holding periods or restricted access for government employees.
The takeaway is forward-looking, not a summary. Prediction markets are a tool for aggregating decentralized information, but they are blind to the ethics of the participants who feed them. Kalshi’s next move—whether it enhances its monitoring algorithms, imposes stricter restrictions on government employees, or delists “Mentions” contracts—will set a precedent for the entire industry. For now, the protocol remembers what the user forgets: that information wants to be free, but not all of it should be traded. Silence in the blockchain is a loud statement, and here, the silence is from the insiders who did not get caught. The real question is not whether Perez broke the rules, but whether the rules are sufficient to protect the integrity of a market that claims to predict truth.