Polymarket Bets on OpenAI, Google Raise Insider Trading Suspicions
The Takeaway
- Polymarkets bets on launches at OpenAI and Google fuel suspicions of insider trading
- Companies are increasingly prohibiting insider trading on prediction markets
- Polymarket, Kalshi volumes have surged in past year
Over the past week, a handful of accounts on Polymarket, the predictions site, bet OpenAI would release a new large language model by December 13. On Thursday December 11, OpenAI released GPT-5.2, and four of these accounts together made over $13,000, according to the trades displayed on their accounts.
The payout is adding fuel to suspicions that a handful of accounts on prediction sites such as Polymarket and Kalshi aren’t just lucky—they’ve got access to private information about tech companies including Google and OpenAI, perhaps because the account owners work there. As the popularity of these prediction sites jumps, more companies are making sure policies that have long prohibited employees from trading stocks based on confidential information also include prediction markets, which allow users to place small bets on events from Taylor Swift’s engagement to the chance of a SpaceX public offering.
KPMG partner Conway Dodge said in the last six months, the number of conversations he’s had with corporate clients about whether their insider trading policies should include prediction markets has at least doubled.
It “might be the next problem that financial institutions and other clients need to start thinking about,” said Dodge, who previously worked in enforcement at the Securities and Exchange Commission.
Crypto and stock trading apps have already recognized the risk. Just over a year ago, Robinhood updated its insider trading policies to apply to prediction markets.
Coinbase at least several months ago expanded its policies to “prohibit employees, including executives, from participating in prediction markets,” a spokesperson for the cryptocurrency exchange said in a statement.
(Robinhood operates its own prediction market and Coinbase is planning to launch prediction markets next week.)
OpenAI and Anthropic, for their part, say their policies clearly restrict employees from using confidential information for personal gain, including wagers on prediction sites. It’s not clear when or whether these policies changed.
The increased attention by companies to employees’ use of prediction sites follows an upswell in activity on Kalshi and Polymarket in the last year after people flocked to bet on the 2024 presidential election. Both sites allow users to purchase event contracts—derivatives that pay out to investors who guess event outcomes correctly—for even less than $1. Users pay for the contracts up front; if they guess correctly, they make the money back plus a profit.
The flexibility to bet on “any difference of opinion,” in the words of Kalshi’s co-founder, has made prediction markets hugely popular. Trading volume on Kalshi, which has touted its oversight by the Commodity Futures Trading Commission, has surged by about five times in the last six months, to an average of $183 million per day over the last seven days, according to data compiled by crypto data provider Artemis Analytics.
In September, Polymarket said the CFTC had given it the go-ahead to serve U.S.-based users after the agency had barred it from accepting those trades three years ago. Its trades have jumped just over six times to an average of $197 million per day. Investors have rushed to back the companies at increasingly high valuations.
As AI has dominated more of the public’s attention, the sites have increasingly offered wagers on tech product releases—which are typically too niche for traditional betting sites. For instance, on Kalshi, users can pay 48 cents to bet that the designer Jony Ive, who currently works with OpenAI, is developing a clip-on device for the company. For 23 cents, they can bet he’s working on a head-mounted display. If those events come true, the contracts are worth $1.
Some users seem clairvoyant, repeatedly making large wagers about the same company in the lead-up to its announcements. That pattern has fueled suspicions that the winning bets are coming from inside the companies.
Last week, a Polymarket account made over $1 million in a day, according to its trade history on the site, with an accurate series of bets about Google’s 2025 search data. This performance raised suspicions among internet commentators that a Google insider was behind the account. A spokesperson for Google declined to comment on whether the company has rules against insider trading on prediction markets.
U.S. securities laws prohibit trading on “material nonpublic information.” But the SEC does not govern prediction markets contracts because they are not securities. Instead, it would be up to the Commodity Futures Trading Commission, which oversees futures trading, or the Department of Justice to pursue such cases, say lawyers.
Still, profiting off confidential information on a prediction market could violate an employee’s legal obligations to their employer. “It’s a form of fraud that’s akin to embezzlement because you’re secretly using the information for your own benefit,” said George Canellos, a lawyer for law firm Milbank LLP who specializes in corporate governance and securities law.
On Thursday, several firms including Kalshi and Coinbase said they formed a new industry group that will advocate for federal rather than state oversight. One of its first initiatives will focus on establishing national standards against insider trading.
Complicating matters is the fact that industry leaders have sometimes suggested there should be room for employees to bet on their own company’s activities. For example, Coinbase CEO Brian Armstrong said he was recently asked whether insider trading should be permitted in prediction markets.
Armstrong responded that it was not “clear-cut,” he recounted at The New York Times DealBook Summit last week. If people wanted to know whether the Suez Canal was going to reopen, he said, the market would be more accurate if an admiral on a ship in the canal was allowed to bet on it. On the other hand, “you want to preserve the integrity of those markets.”
In fact, some companies, including Google and Anthropic, have established their own internal prediction markets. Employees can bet—without using any real real money—on questions such as when a team will finish a project.
In these cases, the markets’ predictions are kept inside the company, so they do not harm the company, said Dan Schwarz, who built Google’s current prediction market and served as chief technology officer of forecasting site Metaculus. For these internal prediction markets, rather than discouraging insiders, “you’re trying to get insider trading,” he said. “You’re trying to get people to reveal what they know.”