White House Teleprompter Operator in Settlement Talks Over Alleged Kalshi Betting

Federal regulators are in settlement talks with Gabriel Perez, President Trump's longtime teleprompter operator, over alleged insider trading on the prediction market platform Kalshi, according to a report published July 16, 2026 by NPR.
Perez, whose official White House title is technical assistant to the president, has worked for Trump since 2016. He is alleged to have made nearly $100,000 on Kalshi by profiting off his access to the president's prepared remarks, according to NPR. The alleged scheme centered on Kalshi's "mention markets," where users bet on specific words or phrases the president will or will not say during public addresses.
The Commodity Futures Trading Commission, the federal agency that oversees Kalshi, is negotiating with Perez over his alleged betting activity. ABC News first reported the investigation. Kalshi's internal surveillance systems detected unusual betting patterns on mention markets involving the president, and investigators determined the account holder was a federal employee. Robert DeNault, head of enforcement at Kalshi, confirmed that the company's surveillance team flagged the trades and referred them to the CFTC.
Kalshi froze approximately $90,000 of Perez's profits and banned him from the platform. The company also announced it is introducing new security measures in response to the investigation, according to ABC News.
This is the first known instance of a White House employee being investigated for allegedly using access to nonpublic government information for prediction market profits, according to NPR.
In March 2026, White House staff received a memo warning against using nonpublic government information to place bets on Kalshi and Polymarket. The memo stated that it is a criminal offense for anyone inside the White House to "buy" or "sell" on prediction market sites, and that misuse of government information "will not be tolerated."
The prediction market activity around presidential remarks has grown substantially. Ahead of Trump's address to the nation on July 16, 2026, Kalshi traders had wagered more than $800,000 on mention markets about what words the president would say.
The broader context here is the intersection of federally regulated prediction markets and the information access held by White House personnel. Prediction markets like Kalshi, which operates under CFTC authority, allow users to bet on political and economic events. The mention market product lets users bet on specific words in public remarks, creating a direct link from nonpublic speech drafts to potential financial gain for anyone with advance access to those drafts.
The legal exposure for Perez is significant. Using nonpublic information to trade on Kalshi could be criminally prosecuted as wire fraud, commodities fraud, and money laundering, according to NPR. The CFTC's interest in a settlement rather than an immediate referral for criminal prosecution suggests the civil enforcement process is still running its course. Whether the Justice Department opens a parallel criminal investigation is a separate question.
The case also tests Kalshi's own compliance infrastructure. The platform's surveillance systems did detect the unusual activity and refer it to the CFTC, and the company acted to freeze funds and ban the user. The new security measures Kalshi says it is implementing in response to the investigation will be closely watched by both regulators and competitors in the prediction market space.
For White House counsel and ethics officials, the Perez matter may sharpen enforcement of existing guidance. The March 2026 memo covering Kalshi and Polymarket was explicit about the criminal stakes and the prohibition on using government information for prediction market gains. The fact that a staff member with direct access to presidential remarks was allegedly able to place bets on the content of those remarks, despite the memo, will raise questions about whether advisory memos are sufficient or whether more active monitoring is needed.


