OpenAI terminated an employee earlier this year for using confidential company information to place bets on prediction market platforms Polymarket and Kalshi, the company confirmed to Wired. Fidji Simo, OpenAI's CEO of Applications, disclosed the firing in an internal message to staff. The employee's name has not been released.
What makes this case worth paying attention to isn't the firing itself. It's how the person got caught, and what the on-chain data suggests about how widespread this might be.
The blockchain receipts
Polymarket runs on the Polygon blockchain, which means every trade is pseudonymous but publicly visible. Analysts at Unusual Whales, a financial data platform, dug through those records and flagged 77 suspected insider positions across 60 wallet addresses tied to OpenAI events going back to March 2023. That's not a rounding error. Sixty wallets.
The clusters tracked major product announcements: Sora, GPT-5, the ChatGPT browser launch, and markets tracking Sam Altman's employment status. One case stands out. In November 2023, two days after Altman was ousted from the company, a brand-new wallet placed a single bet that he'd return. It collected roughly $16,000 in profit and never traded again.
But the pattern that really drew scrutiny was around the ChatGPT browser launch. According to Unusual Whales CEO Matt Saincome, 13 brand-new wallets with zero trading history appeared within 40 hours of the product's public unveiling and collectively bet $309,486 on the correct outcome. All of them fresh. All of them right. That's either the luckiest cohort of crypto degens on the internet, or someone was talking.
How do you catch someone on a pseudonymous blockchain?
This is the part nobody's explaining well. Polymarket wallets don't have names attached to them. You can spin up a new one in seconds. The fired employee reportedly used multiple wallets, which should have made tracing harder. But Unusual Whales weighted accounts by age, trading history, and the size of their bets relative to their activity. New wallets making large, confident, one-time wagers on low-probability events clustered around a single company's announcements? That's a statistical fingerprint.
OpenAI apparently cross-referenced that pattern with internal data: who had access to specific launch dates, who knew about product timelines. Kalshi, unlike Polymarket, requires identity verification, which gave the company another thread to pull. Put the on-chain clustering together with the Kalshi activity records, and one employee matched.
Whether that single termination covers the full scope of the problem is a different question. Saincome told Wired the data points to activity across the tech sector, not just OpenAI. A pseudonymous account dubbed the "Google whale" reportedly made over $1 million trading on Google-related events. Google didn't respond to Wired's inquiries about its policies.
The legal gray zone
OpenAI treated this as insider trading and fired the person for violating internal policy. Spokesperson Kayla Wood confirmed the company prohibits employees from using confidential information for personal gain, including on prediction markets. Straightforward, at least internally.
Legally, it's messier. Traditional insider trading law under the Securities Exchange Act applies to securities. Prediction market contracts aren't classified as securities. On Kalshi, they're regulated as event contracts under the Commodity Exchange Act, where the rules around information asymmetry are less developed. On Polymarket, which operates on blockchain infrastructure, the regulatory framework is thinner still. Someone betting on an upcoming product launch with inside knowledge might not be breaking federal law, even if it looks and smells exactly like the Wall Street version.
OpenAI isn't waiting for Congress to figure this out, though. And they're not alone. The same week the firing was reported, Kalshi disclosed its own enforcement actions: a MrBeast video editor named Artem Kaptur was fined roughly $20,000 and suspended for two years after betting on outcomes of MrBeast videos using information from his job. A California gubernatorial candidate was banned for wagering on his own race. Both cases were referred to the CFTC, which issued an advisory reminding platforms that it retains authority to investigate and prosecute violations on designated contract markets.
What doesn't add up
Seventy-seven flagged positions across 60 wallets over nearly three years, and OpenAI fires one person. Maybe that one person was responsible for all of it, rotating through disposable wallets. Or maybe the company found one clear case and stopped digging. The 13-wallet cluster before the browser launch, in particular, feels like more than one person's operation: $309,486 spread across fresh accounts suggests either someone with a lot of capital and very poor operational security, or multiple people sharing the same tip.
Prediction markets currently handle billions in volume annually. Polymarket alone processed over $3 billion last year. Active wagers on OpenAI's 2026 product roadmap and potential IPO timeline are still live on both platforms. The structural incentive hasn't changed: if you know when a product ships, you can profit. The question is whether a single firing and a few platform bans are enough to change anyone's risk calculus, or whether this is just the cost of doing business on markets that were, until very recently, openly inviting informed traders to participate.
