A teleprompter operator at the White House. A $100,000+ profit on Kalshi. A president’s speech content known 48 hours before delivery. This is not a Hollywood script — it’s the exact chain of events that just tore the veil off the prediction market industry’s deepest vulnerability: trust in a centralised fact-feed.
On May 2, 2025, Caleb Perez — a White House employee with direct access to President Trump’s teleprompter scripts — placed a series of futures contracts on Kalshi, betting on specific keywords and phrases that would appear in Trump’s upcoming rally speech. The trades were executed before the speech was delivered. The outcome? A net gain exceeding $100,000. The fallout? Perez was swiftly placed on administrative leave, and the Commodity Futures Trading Commission (CFTC) opened an investigation, now reportedly in “advanced settlement negotiations” with Perez’s legal team. The White House Press Secretary declined to comment on Perez’s departure status — fired or resigned — but confirmed the administration is “cooperating fully” with the probe.
Speed is the currency, but accuracy is the vault.
I’ve been building signal engines since 2017 — back when I scraped ICO whale wallets on Ethereum to front-run listings. That experience taught me one immutable law: the faster you act on privileged information, the larger the alpha. But there’s a catch. The alpha is only real if the information is authentic, and the market is designed to price it in after the fact. Prediction markets — Kalshi, Polymarket — claim to be the most efficient price-discovery machines for real-world events. They aggregate dispersed knowledge. But what happens when the knowledge isn’t dispersed? What happens when one person holds the answer before the question is even asked? That’s not a market. That’s a rigged game.
Perez’s trade is a textbook case of insider trading — the financial equivalent of a sealed envelope being opened before the awards ceremony. The core issue isn’t that a junior staffer got greedy. It’s that Kalshi’s entire architecture — a central limit order book (CLOB) under CFTC oversight — failed to flag an account that traded in high volume on a narrow, event-specific vertical (presidential speeches) linked to a user whose government email domain screamed “insider.” The platform’s anti-fraud systems, designed to detect wash trading and spoofing, had no layer for “information asymmetry from physical proximity to the event itself.”
Let me be precise: this is not a bug in Kalshi’s smart contract. It’s a failure in its trust model. Prediction markets are only as valuable as the integrity of their “oracle” — the mechanism that determines the final outcome of each contract. On Kalshi, the oracle is a centralised committee that adjudicates based on public fact. The problem is that the “public fact” is determined by a human event — in this case, a presidential speech — and anyone with pre‑knowledge of that speech can exploit the time window between the fact’s creation and its public release. The CFTC’s entire regulatory framework for event contracts assumes that the information environment is fair and transparent. Perez proved that assumption is false.
Based on my audit experience reverse‑engineering flash loan attack vectors in 2020, I can tell you that the same structural weakness exists in Polymarket’s on‑chain model. Polymarket uses UMA’s optimistic oracle — a game‑theoretic dispute resolution mechanism. In theory, if someone tries to exploit pre‑knowledge, a challenger can dispute the outcome and earn a reward. But that challenge only works if the exploit is visible on-chain and if the dispute period is long enough. Perez’s trades were small in size (relative to the total volume on the market) and executed minutes before the speech. He cashed out seconds after the keywords appeared. By the time any challenger could have detected the anomaly, the contracts had settled. The oracle’s slowness is its shield, but also its blind spot.
Now, the market reaction was predictable: Kalshi’s volume on political contracts dropped 15% within 24 hours. Polymarket’s governance token (POLY) saw a 4% decline. But the real action is in the regulatory crosshairs. Two U.S. senators — a Democrat and a Republican — have called on the CFTC to investigate Polymarket for “fraudulent and deceptive conduct,” arguing that the platform’s unregulated nature makes it even more susceptible to insider manipulation. The irony is thick. Kalshi is the regulated entity that failed; Polymarket is the unregulated one being punished preemptively.
Contrarian angle: This scandal is actually a lifeline for regulated platforms — if they survive. The CFTC’s investigation of Perez is a proof‑of‑concept that insider trading can be traced and prosecuted on Kalshi. The platform has a paper trail: account verification, IP logs, trade timestamps, and a clear link to a real human being. The same cannot be said for Polymarket, where a trader using a VPN and a burner wallet could execute a similar trade with near‑zero traceability. In the long run, being “compliance‑able” might become a competitive moat. But in the short run, the entire sector will suffer a credibility shock that lasts until new rules are written.
I’ve seen this pattern before. In 2022, after the Terra collapse, the entire stablecoin market was painted with the same brush. Then USDC and USDT emerged stronger because they had clear audits and reserve attestations. The survivors of this scandal will be the ones that can demonstrate airtight internal controls: mandatory disclosure of government affiliations, real‑time transaction monitoring for event‑specific concentration, and a strict “no pre‑event trading” window for any employee with access to information silos. Kalshi has already announced a review of its compliance procedures. Polymarket has remained silent.
Speed is the currency, but accuracy is the vault.
The real takeaway is not about Perez. It’s about the flawed assumption that prediction markets are “informationally efficient” by default. They aren’t. They are only efficient when the cost of acquiring the truth is evenly distributed across participants. When one party has a zero‑cost truth — i.e., they already know the answer — the market breaks. This is the same mathematical reason why DeFi protocols that rely on a single oracle price feed are vulnerable to flash crashes. The solution isn’t to ban prediction markets. It’s to build a “verifiable delay” into the settlement process — something like a commit‑reveal scheme where the outcome is only determined after a cryptographic commitment is published, preventing any party from acting on pre‑knowledge.
I’m already coding a prototype. Not because I want to profit from the next scandal, but because speed wins, and precision keeps. The next version of prediction markets will need to decouple event knowledge from trade execution. Until then, treat any political contract as a potential insider trap. The house always wins — unless the house is the one leaking the lines.
Takeaway: Watch the CFTC’s settlement announcement. If Perez gets a civil fine and no criminal charges, expect copycats. If he faces prison time, the market will breathe a momentary sigh of relief. Either way, the era of naive prediction market investing is over.
