OpenAI Releases GPT-5.2 One week before the release of GPT-5.2, several Polymarket accounts placed bets that the company would release a new model, profiting over $13,000. Similar suspicious transactions have raised insider trading concerns, prompting tech companies and financial institutions to include prediction markets under regulatory oversight. This article is sourced from Wall Street Insights, written and compiled by Foresight News.
(Background: The third anniversary of ChatGPT: The end of the large model war, where is the true moat?)
(Additional background: ChatGPT launches group chat: up to 20 people can chat simultaneously, Taiwan gets early access (two-step immediate use))

Table of Contents

• Suspicious Transaction Patterns Raise Attention
• The Gray Area of Laws and Regulations
• Industry Leaders’ Contradictory Stances
• Rapid Expansion of Prediction Markets

One week before the release of GPT-5.2, several Polymarket accounts bet that a new large language model would be released before December 13. After the product launch, four accounts collectively profited over $13,000. Also last week, a Polymarket account accurately bet on Google’s 2025 search data, earning over $1 million in a single day. These abnormal account performances are suspected to be operated by insiders, prompting more tech companies and financial institutions to regulate prediction markets as insider trading.

Prediction markets are facing scrutiny over insider trading allegations. According to The Information, several Polymarket accounts recently placed bets and profited before major product releases from OpenAI and Google, raising questions about whether insiders are exploiting these platforms. This has led to increased efforts by tech companies and financial institutions to regulate prediction markets as insider trading.

On December 11, one week before the release of GPT-5.2, several Polymarket accounts bet that a new large language model would be released before December 13. After the product launch, four accounts earned over $13,000. Similarly, last week, a Polymarket account accurately bet on Google’s 2025 search data, making over $1 million in a single day. These suspicious accounts are suspected to be operated by insiders.

This phenomenon is driving policy changes within companies. Conway Dodge, a partner at KPMG, said that discussions with corporate clients about including prediction markets in insider trading policies have doubled over the past six months. Robinhood updated its policies over a year ago to cover prediction markets, and Coinbase expanded its policy months ago to “prohibit employees, including senior executives, from participating in prediction markets.” OpenAI and Anthropic stated that their policies explicitly restrict employees from using confidential information for personal gain, including betting on prediction websites.

The rapid growth of prediction markets is intensifying regulatory urgency. According to data from crypto data provider Artemis Analytics, Kalshi’s trading volume has surged about fivefold over the past six months, with an average daily trading volume of $183 million in the past week. Polymarket’s daily trading volume has increased more than six times to $197 million.

Suspicious Transaction Patterns Raise Attention

Some accounts’ “foresight” has become a focus of suspicion. These users repeatedly place large bets before the same company makes announcements, with extremely high accuracy.

According to transaction records on Polymarket, four accounts betting on OpenAI releasing a new model had already entered the market a week before the product launch. When GPT-5.2 was released as scheduled, these early-positioned accounts quickly cashed out their gains.

Last week, a more controversial case involved a bet on Google. This account made a series of accurate predictions about Google’s 2025 search data, earning over $1 million in one day. This performance has sparked widespread suspicion among online commentators, who believe the account may be operated by Google insiders. Google’s spokesperson declined to comment on whether the company has insider trading regulations related to prediction markets.

As AI garners more public attention, these prediction sites increasingly offer betting options related to tech product launches—topics that are too niche for traditional betting sites. For example, on Kalshi, users can spend 48 cents to bet that designer Jony Ive, currently collaborating with OpenAI, is developing a clip-on device for the company, or 23 cents to bet that he is developing a head-mounted display. If the bet is correct, the contract is worth $1.

Laws and Regulations in the Gray Area

The insider trading issues in prediction markets are in a regulatory blind spot. U.S. securities law prohibits trading based on “material nonpublic information,” but the U.S. Securities and Exchange Commission (SEC) does not regulate prediction market contracts because they are not classified as securities.

Lawyers say jurisdiction over such cases would fall to the Commodity Futures Trading Commission (CFTC), which regulates futures trading, or the Department of Justice. Nonetheless, exploiting confidential information for profit in prediction markets may violate employees’ legal obligations to their employers. George Canellos, a lawyer specializing in corporate governance and securities law at Milbank LLP, said: “This is a form of fraud similar to embezzlement because you are secretly using information for personal gain.”

Dodge, who previously worked in SEC enforcement, believes this “may be the next issue that financial institutions and other clients need to start considering.”

On Thursday, several companies including Kalshi and Coinbase announced the formation of a new industry organization advocating for federal rather than state regulation. One of its initial initiatives will focus on establishing national standards for insider trading.

Industry Leaders’ Contradictory Stances

Adding complexity to the issue, industry leaders sometimes hint that employees should be allowed to bet on their own companies’ activities. Coinbase CEO Brian Armstrong revealed at the New York Times DealBook Summit last week that he was recently asked whether insider trading should be permitted in prediction markets.

Armstrong responded that the question “is not clear.” He gave an example: if people want to know whether the Suez Canal will reopen, allowing naval officers on ships in the canal to participate in bets could make market predictions more accurate. But on the other hand, “you want to maintain the integrity of these markets.”

In fact, some companies, including Google and Anthropic, have already established internal prediction markets. Employees can bet on questions like when their team will complete a project without using real money.

Dan Schwarz, who built Google’s current prediction market and is CTO of the prediction site Metaculus, said that in these cases, the markets are kept internal, so they do not harm the company’s interests. For these internal prediction markets, “you’re not trying to prevent insider trading, but rather trying to facilitate it. You’re trying to get people to reveal what they know.”

Rapid Expansion of Prediction Markets

Concerns over regulation are coinciding with explosive growth in these platforms. After a surge of users betting on the 2024 presidential election, Kalshi and Polymarket have seen activity increase over the past year. Both sites allow users to buy event contracts for less than $1—derivatives that pay out to investors who correctly predict the outcome. Users pay upfront for the contracts, and if they guess correctly, they can recover their funds and make a profit.

Kalshi’s co-founder said that the flexibility to bet “on any disagreement” has made prediction markets very popular. According to data compiled by Artemis Analytics, Kalshi, which operates under CFTC regulation, has seen trading volume increase about fivefold over the past six months, with an average daily trading volume of $183 million in the past week.

In September, Polymarket announced that the CFTC had approved it to offer services to U.S. users, after being banned from accepting such trades three years earlier. Its trading volume has surged over six times to an average of $197 million daily. Investors are increasingly supporting these companies at higher valuations.

Robinhood operates its own prediction market, and Coinbase plans to launch one next week.

!Official website TG banner-1116 | Blockchain News Media - The Most Influential Blockchain News Outlet