OpenAI just fired an employee for a reason that will make your jaw drop. The person was using company secrets to place bets on prediction markets like Polymarket and cashing in big. Even more shocking, the investigation found this was not a one-time thing. Over the past year, 60 mystery wallets made 77 eerily accurate insider bets.
It was late at night in San Francisco. The lights at OpenAI headquarters on Mission Street were still burning bright.
But tonight, what kept the executives awake was not some breakthrough in AGI. It was not the training curves of GPT-6. It was a bizarre internal spy game that no one saw coming.
OpenAI leadership swung the axe and fired an employee on the spot.
The reason was absolutely stunning.
Instead of writing code and tuning parameters for large models, this person secretly took OpenAI’s core secrets and ran to Polymarket to place wild bets, hoping to get rich overnight.
The moment the news broke, all of Silicon Valley exploded.
From WIRED to TechCrunch, from Gizmodo to every major Wall Street outlet, the story spread fast. Everyone was asking the same question.
In the holiest temple of AI research on Earth, how many gambling pros are hiding in the shadows, ready to harvest the entire internet?
our dream ai
One Internal Email Tears Open the Tip of the Iceberg
The story began with an hentai ai chat internal all-hands email from Fidji Simo, OpenAI’s head of applications.
Fidji Simo, a former Meta executive who now steers OpenAI’s application business, sent a deeply serious internal message earlier this year. She told every employee a shocking piece of news.
The company had finished an internal investigation and fired one staff member.
The reason was simple. The employee had used OpenAI’s confidential information girlfriend gpt to trade on external prediction markets. This violated the company’s strict policy against using insider knowledge for personal gain.
The email did not name the employee. It did not mention how much money was involved. It did not say which secrets were leaked.
Everything was buried under a thick layer of silence. But the more they hid, the more people imagined the worst.
WIRED senior writer Kate Knibbs smelled blood in the water first. Her bombshell investigation dropped like a depth charge and blew up across the tech world.
Following that lead, everyone soon realized this was not one person’s crime. It was a planned, organized, and deliberate information arbitrage operation.
13 Ghost Wallets: A 40-Hour Precision Heist
To understand how crazy this insider drama really is, we need to look at a set of numbers that will make your scalp tingle.
The data comes from Unusual Whales, a financial data analysis platform.
This platform does one thing and does it well. Like a shark smelling blood, it tracks every unusual trade across the web.
They zeroed in on every prediction market event tied to OpenAI. Then they found a chilling pattern.
Since March 2023, prediction markets around major OpenAI events have seen a flood of trades that are too accurate to be luck.
The data shows 60 wallet addresses placed 77 suspected insider positions.
These wallets bet on almost every major OpenAI product launch and internal event with scary precision.
When will Sora launch? Bet. Correct.
When will GPT-5 drop? Bet. Correct again.
When will the ChatGPT browser go live? Bet. Still correct.
But the creepiest part was not even those hits.
The truly weird scene happened around the days when the ChatGPT browser made its stunning debut.
The entire tech world was buzzing with hype. Every major outlet was covering it. Prediction markets were flooded with 13 brand new ghost wallets, each one freshly funded with clean money.
These wallets had zero transaction history before this. Not a single prior trade.
They were clearly created for one purpose only. To bet on this one event.
And sure enough, within just 40 hours, those 13 wallets moved like a precision machine. They placed bets, cashed out, and vanished. The total profit was around 300,000 dollars, spread across roughly 2.2 million dollars in bets on whether the ChatGPT browser would launch on time.
The automated tracking systems caught it all.
The pattern was too perfect. The timing was too tight. The wallets were too new. Every single one of them hit the mark.
Grab one accurate bet and you might call it a lucky guess. Grab 77 in a row and it is not luck. It is a leak.
The Sam Altman Firing: Where the Leak Began
Some people say the ChatGPT browser leak was just a company sighing at its own internal chaos. But the real earthquake happened back in November 2023.
That was the weekend when the entire AI world watched in shock as OpenAI’s board suddenly fired CEO Sam Altman.
The whole internet went dark. Investors panicked. Engineers pulled all-nighters. Employees flooded Slack with one shared question. Is the company going to survive?
And then, the bets came.
Unusual Whales data shows that in the 48 hours before the board made its move, a brand new wallet appeared on a prediction market and placed a massive bet on Altman being fired. The wallet had no history. It was created just days before the event.
Sam Altman was fired by the board.
Altman was later reinstated. Microsoft backed him. The chaos ended. But the prediction market had already paid out. That single wallet cashed in roughly 1.6 million dollars.
The house always wins. But in this case, the house was inside the house.
We all thought the people gathered at OpenAI, the brightest minds on Earth, were thinking about how to push humanity past the singularity of AGI. We thought they were using code and computing power to unlock the deepest mysteries of the universe.
The reality? Some of them were just looking for the fastest way to make a buck.
And this is not an isolated case. It is a systemic problem that touches every corner of the tech world.
What Is Polymarket and Why Does It Matter
Before we go further, let us explain what a prediction market is.
Polymarket is the largest decentralized prediction market on the internet. In simple terms, it is a place where you do not bet on sports teams. You bet on real-world events.
Here is an example. Will Trump win the 2024 election? You bet Yes or No. If you are right, you win USDC stablecoins. If you are wrong, you lose your stake.
The logic behind prediction markets is crowd wisdom. When thousands of people put real money on the line, the market price becomes a surprisingly accurate forecast of what will happen.
In a recent 60 Minutes interview, Polymarket CEO Shayne Coplan put it this way. The goal of Polymarket is not gambling. It is to create a tool that uses the wisdom of crowds to give the world better information.
In other words, the platform thrives on people with real information joining the market. That is what makes prices accurate. That is the whole point.
But here is the catch. When an OpenAI employee walks in with absolute insider knowledge, it is no longer a prediction. It is a theft.
Using a prediction market to turn company secrets into personal profit is not clever. It is not a gray area. It is plain insider trading, dressed up in crypto clothing.
The Platforms Are Trapped
The most ironic part is the awkward position prediction market platforms now find themselves in.
Not long ago, these platforms treated information arbitrageurs as honored guests. The CEO of Polymarket publicly welcomed anyone with an information advantage, claiming it makes market prices more accurate and closer to the truth.
But now?
When an OpenAI employee brings company secrets to price discovery, when a MrBeast video editor uses backstage data to optimize information efficiency, when a politician uses absolute knowledge of their own campaign to calibrate probabilities, the platforms suddenly find themselves in a bind.
A Gizmodo report nailed the contradiction. In the short term, inviting insiders to trade helps platforms because information gets sharper. But in the long run, if everyday users know they are gambling against players with classified intel, who will keep playing? And if these platforms start losing corporate partners, that is the real death blow.
That analysis is spot on. OpenAI can fire one mole today. Tomorrow every major tech company will launch internal audits. When your employees turn company strategy into public betting chips on prediction markets, no board can sit still.
Prediction markets went from darling of financial innovation to gray channel for information leaks in just a few steps.
The MrBeast and Kalshi Connection
This is not just an OpenAI problem. The same pattern is showing up everywhere.
MrBeast, the biggest creator on YouTube with over 3 billion fans worldwide, runs a massive content machine. For every video, his team places bets on Kalshi about whether the video will hit certain viewership milestones.
One of his video editors used insider knowledge of upcoming release schedules, topics, and timing to place bets on Kalshi. The platform’s own tracking system flagged the unusual accuracy. The editor cashed in roughly 4,000 dollars per bet. The success rate was so high that Kalshi’s own systems caught it and reported it to authorities.
Kalshi officially stated that the user was flagged for suspicious trading behavior. After a one-year investigation involving over 200 data points, the company concluded that the account showed clear signs of being an information leak.
The Platforms Eat Their Own Words
The irony is thick. Prediction market platforms are now victims of their own success.
They built their brands on the idea that more information makes markets smarter. They welcomed insiders with open arms. The CEO of Polymarket said information advantages make prices more accurate and bring markets closer to truth.
But when the insiders are your own employees, when the information advantage is stolen company property, the whole story falls apart.
OpenAI employees using company secrets to move market prices is not price discovery. It is corporate espionage with a betting slip.
And the damage goes both ways. The employee profits. The platform gets accurate data. But the company loses trust. The public loses faith. And the prediction market itself becomes a laundering machine for stolen information.
We always imagined that the brightest minds gathered at OpenAI were thinking about how to push human civilization past the singularity of AGI. We thought they were using code and computing power to unlock the deepest secrets of the universe.
The reality? Some of them could not resist the temptation to turn the world’s most important secrets into betting chips on a prediction market.
When AI Alignment Meets Human Misalignment
The deepest irony of all is this.
OpenAI has spent billions of dollars and assembled the world’s top scientists to study one question. How do we align AI with human values?
They think about it every day. How do we make sure superintelligence does not betray humanity? How do we build AI systems that are honest, trustworthy, and controllable? How do we stop AI from harming humans in pursuit of its own goals?
And what happened?
AI did not go rogue. Humans did.
You can align a model’s values perfectly and still fail, because the people operating it have a broken moral compass. You can use RLHF to teach GPT what honesty means. But you cannot teach some humans what a底线 means.
That is the most painful part of this whole story.
We have spent years fearing that AI might spin out of control. But we forgot a simpler question.
When humans cannot even control themselves, what gives them the right to align an intelligence more powerful than their own?
