A White House teleprompter operator who worked on President Donald Trump’s prepared remarks allegedly made about $100,000 by trading on whether particular words and phrases would appear in Trump speeches, according to reports from NPR and ABC News.
The aide, Gabriel Perez, had access to a very useful edge: the script. NPR reported, citing sources, that Perez traded on Kalshi “mention market” contracts over several months late last year and early this year. These contracts let users put money on whether specified words will be said in a defined setting, such as a corporate earnings call or, in this case, presidential remarks.
That is the whole mechanism. A trader buys a contract tied to a word or phrase. If the word appears under the market’s rules, the contract pays out. If it does not, it loses. For a person with advance access to prepared presidential text, that stops looking like public forecasting and starts looking like a very dumb way to leave an audit trail.
ABC News reported, citing sources, that Perez often had the last review of Trump’s prepared remarks and sometimes received late edits from Trump himself. Investigators found instances in which Perez allegedly exited bets during speeches when Trump skipped sections containing words Perez had wagered would be spoken, ABC reported.
Kalshi flagged unusual activity, investigated, determined the customer was a federal employee, froze the funds and referred the matter to the Commodity Futures Trading Commission, according to the reports. The CFTC is said to have investigated and to be in settlement discussions with Perez. ABC reported that the CFTC notified federal prosecutors in Manhattan, who declined to open a criminal investigation.
The White House told the Washington Examiner that Perez “will no longer be working at the White House.” CBS News reported in March that the White House had warned staff not to use nonpublic information to buy or sell prediction-market contracts.
Prediction markets meet government access
Kalshi presents itself as a federally regulated prediction market rather than a gambling site. The company has argued that its contracts are closer to commodity futures than casino bets, and it has sought oversight through the CFTC. That federal posture also puts Kalshi and the CFTC in conflict with states that have tried to regulate the products under gambling laws. The CFTC has sued Kentucky, Minnesota, Illinois and Rhode Island, seeking to preempt state rules in favor of a national standard under the agency’s control.
The Perez allegations land in a growing pile of insider-use cases around prediction markets. A U.S. soldier, Gannon Ken Van Dyke, was arrested in April after allegedly using knowledge from planning related to the capture of Venezuela’s Nicolás Maduro to make $410,000 on Polymarket. NPR also reported that former Rep. George Santos faced scrutiny over an alleged Kalshi bet tied to whether he would attend the State of the Union.
The pitch for prediction markets is that prices aggregate information. The problem, shown rather neatly here if the reports are accurate, is that some information is nonpublic because a government employee got it at work. When the market is asking whether the president will say a word, the person loading the teleprompter is not another retail trader with a hunch.
This story draws on original reporting from Ars Technica.
