On Tuesday, July 28, 2026, the New York Times reported that Kalshi and Polymarket are running active prediction markets on drug approval decisions by the Food and Drug Administration. Kalshi has announced plans to go further: it intends to open betting on the results of clinical trials themselves. The story does not say who, specifically, placed the first trade. It does not need to. The architecture is the thing.
A prediction market prices belief. That is the theory, and for elections and commodity prices the theory has some warrant. You want to know what the crowd thinks will happen, you look at what the crowd is willing to lose money on. Fine. But a clinical trial is not an event the crowd is watching from the stands. It is a process performed by a few hundred people — investigators, data managers, statisticians, sponsors, contract research employees — who handle the actual numbers before the actual numbers are public. Betting on the outcome of a process you can corrupt is not forecasting. It has another name.
The trial for a single late-stage cancer drug can run to $300 million or more in development costs and shift a company's market capitalization by billions on the day results drop. The prediction market adds a third ledger: the side bet. It does not require a person to commit outright fraud. It only requires a person to adjust one endpoint slightly, to reclassify one adverse event, to let a borderline result tip one way rather than the other, knowing that the tip has a price attached to it. The incentive does not announce itself. It sits in the room during the analysis meeting.
The FDA approval market is the entry point. Approval decisions are at least made by a government body with a formal record. Kalshi's clinical trial market skips that buffer entirely. It puts a live price on the moment when the data are most malleable and least public.
None of this is hidden. Kalshi states its plans. Polymarket lists its drug markets openly. The concern is not that someone is operating in secret. The concern is that the mechanism is being built in plain view, and the people building it are describing it as a public good — a way to aggregate information, to signal where science is heading. The patients enrolled in the trials are not mentioned in that description. They gave their bodies to the process. They are not listed as participants in the market.
A Phase III oncology trial enrolls a median of 312 patients, runs 36 to 48 months, and asks those patients to accept randomization, side effects, and uncertainty on the premise that the data collected will be handled with care. That premise is now worth something to a third party who never signed the consent form.