The suspension of White House technical assistant Gabriel Perez over allegations of exploiting non-public presidential speech drafts to trade on Kalshi highlights a fundamental vulnerability in the architecture of modern prediction markets. By utilizing early access to the president’s prepared remarks to execute trades in "Mentions" markets—contracts settling on whether specific words or phrases are spoken aloud during public addresses—Perez allegedly extracted over $100,000 in profits. This incident is not merely an ethical breach; it is a structural demonstration of asymmetric information transmission, latency advantages, and the unique risk profile of event-contract design.
Understanding this case requires analyzing the physical flow of information, the mechanics of market execution, and the regulatory boundaries governing prediction market integrity.
The Information Pipeline and Structural Asymmetry
In financial markets, insider trading relies on accessing corporate disclosures before public distribution. In political prediction markets, the equivalent commodity is the speech draft. To understand how the teleprompter operator held an insurmountable edge, we must map the production line of a presidential speech.
[Drafting Cohort] ──> [Presidential Review / Edits] ──> [Teleprompter Loading] ──> [Public Delivery]
│ │
(Gabriel Perez Access Point) (Public Trade Window Closes)
The speech pipeline operates as a series of narrowing funnels:
- The Drafting Stage: A broad team of speechwriters, policy advisers, and agency leads contribute to initial drafts. Information is dispersed and highly volatile, making trading at this stage highly risky due to continuous revisions.
- The Presidential Review: The draft is condensed. The president makes personal, last-minute hand-written edits. The circle of individuals with access to this final version shrinks to a handful of senior aides.
- The Teleprompter Loading Stage: The technical operator receives the finalized text file to load into the software. This file represents the absolute highest-probability state of what will be spoken.
- The Execution Window: Because the operator is physically present at the podium during the live address, they possess a zero-latency feedback loop. They can observe real-time deviations from the script and adjust market positions accordingly.
Perez capitalized on the transition between the teleprompter loading stage and the execution window. By receiving the definitive, edited text immediately prior to airtime, he bypassed the volatility of early drafts. The information asymmetry here was total: the public traded based on historical rhetoric and political intuition, while the operator traded on the exact digital file queued for projection.
Execution Hedging and the Real-Time Correction Loop
The defining feature of this scheme was not just the pre-speech positioning, but the active risk management executed during live broadcasts. Presidential speeches—especially those delivered by Donald Trump—are notorious for rhetorical drift, ad-libbing, and spontaneous omissions.
A standard insider trader in equities buys or sells a position and must wait for public disclosure to realize value, bearing the risk of market misinterpretation. The teleprompter operator, however, possessed a dynamic hedging mechanism.
According to regulatory findings, Perez monitored the live speech text against his open Kalshi contracts. If the president skipped a paragraph or omitted a target word that had been queued in the system, Perez reacted instantly. He initiated mid-speech trades to exit or reverse his positions before the broader market—which relied on video broadcast feeds with inherent streaming latencies of 5 to 30 seconds—could process the omission.
This created an execution model characterized by two distinct phases:
Phase 1: Pre-Speech Positioning (High Alpha)
- Action: Buying underpriced "Yes" contracts for highly specific, unusual terms that were confirmed in the final loaded draft.
- Risk: Low, conditional on the president sticking strictly to the script.
Phase 2: Mid-Speech Real-Time Rebalancing (Risk Mitigation)
- Action: Selling down positions or buying "No" contracts the moment the speaker physically bypassed a section of the text on screen.
- Risk: Zero, as the operator operated with a structural latency advantage over exchange participants watching the broadcast.
Market Surveillance and the Detection Mechanism
The exploitation of this loophole was ultimately checked not by White House internal controls, but by the market design of the prediction exchange itself. Kalshi, as a Commodity Futures Trading Commission (CFTC) regulated entity, operates with mandatory surveillance protocols.
The trades triggered anomalies in Kalshi’s surveillance systems through three main behavioral signatures:
- Abnormal Capital Concentration: Large, directional wagers on illiquid, highly specific "Mentions" contracts shortly before speech times.
- Improbable Win Rates: A consistent, mathematically anomalous profitable streak across more than a dozen highly volatile speech events.
- Synchronized Execution Latency: Trading actions that perfectly correlated with the live, second-by-second teleprompter scroll, showing execution times that preceded broadcast audio delivery to public feeds.
When market makers and Kalshi's internal compliance analysts flagged these patterns, the platform initiated an internal investigation, froze the associated trading account containing approximately $90,000 in undistributed profits, and referred the case to the CFTC.
The regulatory response highlights an ongoing jurisdictional debate. While the Southern District of New York (SDNY) declined to pursue criminal prosecution—likely due to the novelty of applying traditional wire fraud or insider trading statutes to political speech prediction contracts—the CFTC is pursuing a civil settlement focusing on disgorgement of gains.
Structural Vulnerabilities of Mention Markets
The teleprompter exploit exposes a broader design flaw in "Mention" and "Topic" contracts compared to macro-level event contracts (e.g., election outcomes or GDP prints).
Macro-level contracts depend on massive, decentralized data inputs. No single individual can guarantee an employment print or an election result. "Mention" contracts, by contrast, are micro-event contracts where the outcome is entirely controlled by a single individual and processed by a tiny technical team.
| Contract Attribute | Macro Event Contracts (e.g., Inflation, Elections) | Micro Mention Contracts (e.g., Speech Content) |
|---|---|---|
| Outcome Determiner | Distributed systems, millions of participants | Single individual (The Speaker) |
| Information Custodians | Highly secure, multi-agency strict lock-ups | Technical aides, audiovisual staff, transcriptionists |
| Liquidity Profile | High liquidity, tight spreads | Low liquidity, highly vulnerable to manipulation |
| Hedging Capability | Impossible to perfectly hedge via direct action | Easily hedged in real-time by staff |
Because "Mention" contracts feature extremely low liquidity and wide spreads, even small, informed trades can distort the market. This is why major brokerages, such as Robinhood, have intentionally excluded speech-based mention contracts from their prediction offerings, citing the high risk of manipulation and insider exploitation.
Strategic Mitigations for Prediction Exchanges
To preserve market integrity and prevent similar exploits, prediction exchanges and regulatory bodies must shift from reactive surveillance to proactive structural design. Three specific interventions would neutralize the teleprompter loophole:
First, exchanges should enforce a hard trading freeze on all "Mention" contracts 30 minutes prior to the scheduled start of any address. This matches the security standards of traditional financial markets during corporate earnings lockups, rendering last-minute teleprompter adjustments untradable.
Second, exchanges must mandate strict identity and employment disclosures for users trading in politically sensitive categories. Kalshi's move to require workplace disclosures for high-risk markets must be expanded to run automated cross-references against government staff directories and contractor rosters.
Third, market settlement parameters must be adjusted. Contracts should settle based on official, post-event transcripts compiled by neutral third parties, rather than live auditory observation. This eliminates the latency-arbitrage advantage that on-site staff exploit by trading during the speech itself.
The teleprompter incident serves as a warning for the prediction market sector. As these platforms attract larger pools of capital, the incentive to exploit technical roles for information arbitrage will scale proportionally. Without structural adjustments to contract execution windows and strict professional restrictions on technical support staff, micro-event contracts will continue to be compromised by those holding the keyboards.
.jpg?width=1200&auto=webp&trim=593%2C249%2C275%2C1053)
