Nope. In not even an hour and a half, they ruled that Elon Musk had filed his lawsuit too late. The judge threw out the rest of the case as a result. Time to go home, everyone. Thanks for your service.

I’ve been covering this trial from inside the Oakland courthouse since it began — standing in line at 5:30 a.m. for one of the thirty public seats, watching Musk testify about his efforts to take control of OpenAI even as he says he believed (and perhaps hoped) it would change the world, sitting there when Sam Altman was asked within the first two minutes of cross-examination, “Do you tell lies?” It wasn’t just about the billionaire feud, though. It was a rare chance for reporters to see inside the companies, and inside the communications of the tiny handful of people determining the future for all of us.

First, let’s get reacquainted with the case.

Musk sued Altman and OpenAI in February 2024. He put $38 million into OpenAI’s early years on the understanding, he said, that the company would remain a nonprofit dedicated to building AI safely for humanity. When OpenAI created a for-profit subsidiary in 2019, and then struck a $10 billion deal with Microsoft in 2023, Musk said the charity had been looted. He wanted $134 billion in “ill-gotten gains” returned to the OpenAI foundation, Altman and Brockman removed from leadership, and the for-profit restructuring unwound.

The case rested on documents that had never been public before — emails and texts and Greg Brockman’s personal diary entries, subpoenaed during discovery. As I reported, those documents showed the people at the top of the AI industry writing to each other with a casual assumption of world-historical importance. In one email, Musk mused about whether OpenAI would need a twelve-person board or a sixteen-person board — the difference being how much of the fate of humanity was riding on the company.

The testimony that followed was equally illuminating. Ilya Sutskever — the researcher most responsible for what generative AI can actually do — confirmed under oath that he spent a year building a 52-page dossier documenting Altman’s “consistent pattern of lying.” Former CTO Mira Murati, who determined how the technology feels to use, accused Altman of dishonesty. Former board members Helen Toner and Tasha McCauley said the same. Dario Amodei, who left OpenAI to found Anthropic, had said that Altman had misrepresented the terms of the Microsoft investment. As I wrote at the time, the people who worked most closely with Altman described the same pattern under oath. Altman himself was forced to admit that he has lied in his life, sure, although he denied doing it for business advantage.

The jury decided that none of that was the question in front of them. The question was: did Musk know about OpenAI’s shift toward profit-seeking in time to file a lawsuit, and did he wait too long? The jury found he did. Judge Gonzalez Rogers agreed, saying “there’s a substantial amount of evidence to support the jury’s finding.” (The jury’s verdict was to be only advisory, and Judge Gonzalez Rogers would have been the one making the final call under enormous scrutiny — its decision must have come as some relief to her.)

Musk’s team says they’ll appeal. His attorney Steven Molo told reporters the verdict was a narrow decision on “technical legal issues” and that Musk’s side had proved the core of its case. Musk, who skipped the end of the trial for his trip to China with President Trump, posted that Judge Gonzalez Rogers is a “terrible activist” judge and wrote that she “just handed out a free license to loot charities if you can keep the looting quiet for a few years.”

OpenAI’s attorney William Savitt — a very good lawyer who deserves his professional satisfaction today — told reporters outside the courthouse that this was “a hypocritical attempt to sabotage a competitor.”

Both of those framings are, in their way, accurate. And neither of them addresses the more durable problem.

Read more

Paid subscribers get early access to this and all my analysis, as well as written reports, including all the source documents.

So why get on the plane?

When China joined the World Trade Organization in December 2001, it transformed global business. The deal — backed by the Clinton administration, and celebrated (fists pumping, corks popping) in the boardrooms of American technology companies — promised mutual benefit through interconnection. Open markets, shared infrastructure, growth without friction. The companies that would one day be known as Big Tech were barely large enough to understand what it was about to do for them. But what they were building — platforms, supply chains, chip architectures — turned out to be perfect for the world the WTO agreement was creating.

That world lasted about twenty years.

On Wednesday, Air Force One touched down in a new world. It landed at Beijing Capital International Airport carrying President Trump, Secretary of State Marco Rubio, Defense Secretary Pete Hegseth, U.S. Trade Representative Jamieson Greer — and the CEOs of Apple, Nvidia, and Tesla.

Tim Cook, Elon Musk, and Jensen Huang joined Trump’s first visit to Beijing since 2017 alongside the CEOs of Blackrock, Citi, Meta, and others. (Huang joined as a last-minute addition in Alaska, after Trump called to get him literally on board — Pull over, we gotta grab Jensen.) Musk even brought his son X Æ A-XII to the Great Hall of the People.

Quick aside: Dina Powell McCormick, President of Meta (and one of only two women on the trip) is notable in that she’s there in place of Mark Zuckerberg, whose desperate desire to get his products into China once led him to offer Xi Jinping the chance to name Zuckerberg’s son. (Weird offer. Xi refused.) The American uber-capitalists’ pilgrimage to China is a well-worn one.

At a state banquet Wednesday evening, Xi Jinping told the assembled American executives that China’s door would “open wider,” according to state-backed media Xinhua, observing that U.S. companies are “deeply involved in China’s reform and opening up, and both sides have benefited.”

That’s a gracious way of putting it. It also fails to describe the pickle these three CEOs, and the rest of the American tech sector, are in.

The three men who flew to Beijing aren’t looking to expand their business in China. They’re trying to keep from losing it.

Read more

Read more

Subscribe now

A mishmash of lawyers, reporters, and a few protesters watched as Sam Altman was finally sworn in Tuesday morning in a federal courtroom in Oakland to defend himself against accusations of tricking the world’s richest man into funding a sneaky bid to create the world’s most powerful company. I was there.

The case, in its narrowest legal form, asks whether Altman and OpenAI breached a charitable trust by converting a nonprofit into a profit-seeking enterprise. Musk says they promised him something different. Altman says they didn’t.

But as I explained to a half-dozen CNN producers today texting me to see whether they should put this stuff on the air, the testimony wasn’t really about contract law, at least not for me. It was an anthropological exhibit, a power trip, a ringside seat to the billionaire circus. (The press conference on the courthouse steps was framed by a half-dozen elderly women singing “We’ll Stop A.I., Hallelujah,” which I loved, and I’ve discovered that my kink is watching billionaires say “yes, your honor” to Judge Yvette Gonzalez-Rogers.) Two of the most powerful men in technology, both convinced they’re saving civilization, are in court over a falling-out that’s been building for nearly a decade — and in the process, they’re giving us all a portrait of what it looks like when very, very rich men who claim to be steering the future have to account for themselves.

Here’s what stood out.

1. Molo opened the cross-examination with one question: “Are you completely trustworthy?”

Steven Molo is Musk’s attorney, and the dude is not subtle. After a mannered and pleasant direct examination of Altman by his own lawyer, William Savitt, Molo came right in with the big questions: is Altman trustworthy? (You can read about all the people he’s worked with who say otherwise here.) Within the first two minutes of cross-examination, right after exchanging hellos, he asked Altman this:

Molo: Are you completely trustworthy?

Altman: I believe so.

Molo: Do you always tell the truth?

Altman: I believe I’m a truthful person.

Molo: That wasn’t my question, sir. Do you always tell the truth?

Altman: I’m sure there is some time in my life when I have not.

Molo: Do you tell lies to advance your business interests?

Altman: No. No.

Molo: Have you misled people with whom you do business?

Altman: I believe I am an honest and trustworthy business person.

Molo: That wasn’t my question about what you believe. Have you misled people with whom you do business?

Altman: I do not think so.

Molo: Would they think so?

Altman: I can’t answer that for other people.

Gotta hand it to him: Altman was calm and pleasant throughout, while Molo became (or at least acted) angrier and angrier. It wasn’t clear to me whether Molo was honestly irritated, but he couldn’t seem to make Altman angry. Altman maintained the affect of a man who has been asked tougher questions than this. It made me wonder what it would take to accuse Altman of something he’d take offense to.

2. Then Molo read him the list.

What followed was a methodical recitation of every person who has, under oath or on record, called Sam Altman dishonest. It took a while.

Ilya Sutskever, OpenAI co-founder, testified that Altman exhibited “a pattern of lying.” Mira Murati, former CTO, accused him of dishonesty. Helen Toner, former board member, accused him of lying about OpenAI’s safety processes. Tasha McCauley, also a former board member, testified to “a toxic culture of lying.” Soo Yoon accused him of “a lack of concern for the consequences of misleading others.” Dario Amodei accused him of misrepresenting the Microsoft investment terms.

Altman’s answer to most of these was a variation on: I didn’t hear that testimony. Because, after all, as CEO of an $850 billion company, he has other things going on. He couldn’t be in court every day.

Molo: Is it important to you to find out what’s going on in this trial?

Altman: Yes. Although I also have a very busy day job and have not been able to be here every day.

On Dario Amodei specifically, he offered the sharpest line of the day, as close to admitting he doesn’t like someone as I could imagine him saying:

“Dario accuses me of many things.”

3. Altman was entertaining mind-blowingly casual political ambitions in 2017.

While negotiating who would control the technology that might one day produce artificial general intelligence, Altman’s own colleagues — Greg Brockman and Ilya Sutskever — sent him an email asking why the CEO title mattered so much to him. Their email asked, specifically, about his “political goals.”

Molo asked if Altman had told them he wanted to be president of the United States.

Altman: I was thinking about running for governor at the time, which I believe is what they meant here.

Chill, bro. Just governor.

4. The warmest memory he had of Elon Musk was a late-night meeting in 2018 that ended with watching Musk scroll his phone.

Savitt, on direct examination, was walking Altman through the 2018 period when relations with Musk were still, relatively speaking, functional. Altman described a meeting at Tesla headquarters — about the structure of what would become OpenAI’s for-profit entity, the future of AI governance, billions of dollars — and how it had gone unusually well.

“This was, like, a pretty late night meeting, and then it was a long conversation about him showing us memes on his phone. It was just like him showing us memes on his phone. So that was, like, the happiest and calm part.”

Ah, the good times.

5. Shivon Zilis remaining on the OpenAI board after he discovered she was the mother of Musk’s children was, in Altman’s words, “a close call.”

Read more

Ilya Sutskever — the AI researcher who helped build OpenAI and is widely credited with turning the transformer model into ChatGPT — took the stand on Monday and confirmed under oath that he spent a year assembling a 52-page dossier on Sam Altman’s conduct, concluding that Altman “exhibits a consistent pattern of lying, undermining his execs, and pitting his execs against one another.”

That testimony follows sworn statements from former CTO Mira Murati, former board member Helen Toner, and former board member Tasha McCauley — all describing the same pattern. Altman is expected to take the stand today with all of that hanging over him.

But here’s the thing: none of that may actually matter for the outcome of this case. I break down what the jury is actually being asked to decide, why Musk’s legal hill is steeper than it looks, and what Sutskever’s extraordinary testimony reveals about the people building the most consequential technology in the world.

Sam Altman has spent the last two weeks sitting at the defendant’s table in a federal courthouse in Oakland while, one by one, the people who know him best described him to a jury.

Mira Murati, the former CTO who served as OpenAI’s interim CEO during the chaotic week of November 2023, gave video testimony on May 6. She described Altman as someone who was “not always candid,” who “pitted OpenAI executives against each other,” who “undermined” her authority — and whose defining management trait was telling people exactly what they wanted to hear. “My concern was about Sam saying one thing to one person and completely the opposite to another person,” she told the court.

Yesterday, on Day 10, Ilya Sutskever, the tragic zealot of the company, took it further.

Sutskever, a founder and former chief scientist of OpenAI, testified that he had spent about a year gathering evidence for the board. He confirmed that he had prepared a document — 52 pages, according to his prior deposition — at the board’s request, summarizing what he had observed: Altman “exhibits a consistent pattern of lying, undermining his execs, and pitting his execs against one another.” He said he had discussed removing Altman with Murati “for a long time.”

This is not ambiguous testimony. Sutskever was Altman’s ally, his co-founder in effect, one of the most celebrated AI researchers in the world. He has swerved back and forth on Altman. He helped fire Altman in November 2023, then lost his nerve and voted to reinstate him, then eventually resigned. His current stake in OpenAI is worth approximately $7 billion. But he still went to federal court on Monday to testify that the man running the most powerful AI company on earth has a “consistent pattern of lying.”

Before Sutskever, the jury heard from Helen Toner, a former oversight board member, who testified that OpenAI’s self-governance failed because of “profit and personal gain” and “the enormous amount of power” Altman would wield if the company succeeded. And from Tasha McCauley, another former board member, who said “a pattern of dishonesty was a very difficult component of Sam’s leadership.”

Today, Sam Altman takes the stand, and tries to project a different image.

Here’s what makes this case genuinely hard to call — and why the jury’s job is harder than the headlines suggest.

The case Musk is making is not about Altman’s character. It’s about a legal theory: that OpenAI breached its charitable mission by converting to a for-profit structure, and that Altman and Brockman deceived Musk into funding that conversion. Character testimony about lying and manipulation is relevant as background — it helps explain why the board moved against Altman — but the jury decides if the defendants are liable and how Musk should be compensated, while Judge Yvonne Gonzalez Rogers makes the final call on any remedies. The jury’s verdict is advisory.

So Altman’s path on the stand is actually clearer than the testimony record suggests. He doesn’t need to rehabilitate his reputation as a manager. He needs to answer one question cleanly: did he deceive Musk about the for-profit conversion, or did Musk know what was happening and choose to leave anyway?

OpenAI’s answer, delivered in opening statements by lead counsel William Savitt, is a clean counter-narrative: “We are here because Mr. Musk didn’t get his way at OpenAI. He quit, saying they would fail for sure. But my clients had the nerve to go on and succeed without him.”

That’s a pretty convincing angle. And the character testimony, damaging as it is, cuts in an unexpected direction: if Altman is the kind of leader his own colleagues describe — managing up, managing sideways, telling each person what they want to hear — then his conduct toward Musk in OpenAI’s early years starts to look less like fraud and more like the same behavior everyone around him experienced. Because in this weird AI hyperscaler world, manipulation is not the same as breach of charitable trust.

What Altman needs to do today is stay specific, stay calm, and make the case that whatever his management style, Musk understood the financial realities of building frontier AI and made his own choices. The trial record gives him some material to work with: both Musk and Altman have said they wanted OpenAI to safely develop AGI for the benefit of humanity and not for any one person’s gain or under any one person’s control. The fight is about who moved first, and, by extension, who gets the credit.

The deeper problem for Altman looks like it would be the jury. Maybe the public record now being assembled in federal court. The board and the executives’ documented warnings, the 52-page memo sent by Altman’s most cherished hire — all of it is now permanent in the public eye. But the system Altman built doesn’t necessarily need its board members to trust him any further than it takes to keep growing.

Tasha McCauley said a pattern of dishonesty was “a very difficult component” of his leadership. The company Altman runs is currently valued at over $850 billion.

Further Reading

• Sutskever testimony, Reuters — Yesterday’s live reporting on Sutskever’s testimony about the “consistent pattern of lying” and the year-long evidence gathering

• Week 1 courtroom dispatch, MIT Technology Review — Trial context, jury structure, Musk’s demeanor under cross

• Musk week 1 testimony, CNBC — Musk’s claims, OpenAI’s counter-narrative, the $38M donation and $150B damages ask

• An amazingly helpful Wiki of the trial, with transcripts, highlights, themes, etc.

That’s the kind of scientist Ilya Sutskever is. The one who sees it first. The one everyone seems to want to be.

Google hired him immediately after AlexNet. Then, in 2015, Elon Musk and Sam Altman set out to recruit him away for their weird AI nonprofit. Musk later said the fight for Sutskever was one of the hardest recruiting battles he’d ever fought — a protracted back-and-forth with Google DeepMind’s Demis Hassabis on one side and Musk on the other. Sutskever went back and forth, agreed to leave, changed his mind, agreed again. Eventually he walked away from a reported $6 million annual salary at Google to become a co-founder and chief scientist at a nonprofit research lab that would pay him $1.9 million and offer no equity. He said later he chose OpenAI for altruistic reasons. Nobody who knew him seemed surprised.

For the next nine years, Sutskever was the technical engine of everything OpenAI built — including ChatGPT, the product that turned a research nonprofit into a $300 billion company. Musk, who is now suing to destroy that company, described Sutskever in a CNBC interview as the “linchpin” of OpenAI’s success. And on Monday he took the stand in Oakland.

The trial now underway in Oakland has introduced the public to a specific cast of characters from the origin story of commercial AI, and they click into a series of Silicon Valley types. Musk is a maximalist who funded OpenAI to counter Google’s dominance in AI, demanded majority control when the nonprofit structure inconvenienced him, and is now suing the organization he left rather than accept that it exists without him. Altman is a builder of institutions — restless, commercially acute, and effective in the way that people who are comfortable with contradiction tend to be effective. Mira Murati, OpenAI’s former CTO — the person most responsible for what generative AI actually feels like to use — is the pragmatic builder who found herself swerving to do whatever would save the company. She provided Sutskever the screenshots he used to justify firing Altman — and then, in real time as interim CEO, helped Altman engineer his own reinstatement. These are people possessed of the typical forms of ambition: competitive, acquisitive, institutional.

Sutskever is something different, and even his adversaries in this trial seem to feel it.

Read more

The question was this: how can a man with a nearly $30 billion stake in a company claim that the benefit of humanity, and not money, is driving him?

OpenAI President Greg Brockman’s answer, each time, was some version of the same thing: the money was never the point. The mission drove every decision. The for-profit structure was in service of the nonprofit’s goals. But in federal court in Oakland, Molo tried to to jam him with a sentence Brockman wrote in his journal back in 2017:

“Financially, what will take me to $1B?”

Brockman wrote that while OpenAI was still accepting charitable donations from Elon Musk. He wrote it while publicly committed to the nonprofit mission. But today he insists that the financial opportunity and the world-changing possibilities coexisted in his thinking without contradiction, and that through it all, in spite of idly jotting down his thoughts about the path to a billion dollars, money was not his primary motivation. Most people who’ve never written a sentence like Brockman did might find becoming a billionaire and serving humanity irreconcilable. And Molo tried to put that tension in the starkest possible terms for a jury:

“It takes $30 billion to get you out of bed in the morning, but $1 billion doesn’t?”

“That’s not what I’m saying,” Brockman replied.

The question of what he is saying — what anyone in the 0.01% is saying when they insist the money is incidental — is worth exploring. Because I don’t believe it’s a lie. It is instead the worldview of people who have been so thoroughly reshaped by wealth that they’ve lost the ability to see it the way you and I do.

To understand how someone worth $30 billion can mean it when they say money isn’t their primary motivation, you need to understand how equity works in practice — not just as a financial instrument, but as a psychological one.

Read more

I spent two days inside the federal courthouse in Oakland, California, watching Elon Musk get examined and cross-examined in his lawsuit against Sam Altman and OpenAI. The case turns on a founding betrayal — Musk put in $38 million on the understanding that OpenAI would stay a nonprofit, and discovery documents suggest the conversion to for-profit was already being planned while he was still writing checks.

But this trial is about more than one lawsuit. Congress investigated AI and passed no regulations. The FTC looked at AI market concentration and retreated. The California AG signed off on OpenAI’s nonprofit conversion without a fight. The fastest-growing company in corporate history is now having its fate decided by nine jurors drawn from Oakland’s voter rolls — and that jury may be the only formal accountability mechanism the AI industry has ever faced.

In this video: what it was like to be in the room, what the Brockman diary entries actually show, what Musk’s attorneys are really arguing, and why this trial matters far beyond the $38 million at stake.

Further Reading

• Musk v. Altman, Case Filing Documents — Primary court filings including founding emails and Brockman diary entries entered into evidence

• Karim Nader’s reconsolidation research, McGill University — Original science on memory malleability and the reconsolidation process

• Elizabeth Loftus, “Creating False Memories,” Scientific American — Foundational work on memory distortion and false recall

• California AG’s approval of OpenAI’s nonprofit conversion — The regulatory moment that left no footprint

• OpenAI’s corporate restructuring announcement, 2024 — The company’s own characterization of the for-profit shift

I attended the proceedings not by virtue of my new role at CNN, not because I knew most of the reporters and all but one of the camera crews there, but because I brought a driver’s license and showed up early enough (5:32am) to get one of 30 spots in line for anyone who wants one. On one side of me in line in the dark was the Washington Post’s A.I. reporter. On the other side was Janet, who had driven in from Walnut Creek because she just wanted to watch. I went through security just before OpenAI President Greg Brockman, who had to put his phone and wallet through the same machine I did. When I arrived upstairs and was asked who I was with, I simply said “I’m a member of the public,” and was ushered through the doors. U.S.A.!

So what did I witness, there in the room with Elon Musk on the stand, with Sam Altman huddled with his lawyers, with the soft clack of two dozen journalists typing as fast as they could?

Sensitivity. Runaway pride. The deep, unquenchable thirst for credit and legacy that being the leaders (figureheads?) atop a new industry seems to produce. And a strange mix of corporate jargon, creative accounting, and ruler-of-the-world language that veers between admirable pluck and dystopian satire.

Here’s an example:

Read more

In 2000, a neuroscientist named Karim Nader published a paper in Nature that blew up his field.

For nearly a century, the prevailing dogma was that memory worked in two stages. A fresh memory was labile — soft, susceptible to disruption. Then, over hours, it consolidated into stable long-term storage. Once consolidated, it was fixed. You could retrieve it as many times as you wanted. It would always say the same thing.

Nader’s experiment showed this. He fear-conditioned rats — pairing a tone with a mild shock until the sound alone made them freeze — then waited a day for the memory to consolidate. He played the tone once more, to reactivate the memory, and immediately after, he injected a protein-synthesis inhibitor called anisomycin directly into the amygdala to interrupt the process. The result? He found that the rats’ fear was gone. Not suppressed. Gone.

The implication was structural: every time you retrieve a memory, it briefly returns to a labile state and has to be actively rewritten. Retrieval isn’t just reading. It’s read-write access. Every time you reach back for something you know, that memory is in play again — and whatever is present at that moment, all your current fears and needs and interests, gets folded in before it restabilizes. The act of remembering turns out to be an act of editing.

Nader called this reconsolidation. It is now among the most replicated findings in memory research.

This week, Elon Musk took the stand in Oakland and reconsolidated his memory of what he and Sam Altman agreed to in 2015. As I’ve been watching today, he says he and Altman had in mind a nonprofit that would literally be in charge of ushering in a new era for humanity. The texts and emails shown in court have some pretty casual ruler-of-worlds language. Stuff like this:

Read more

Read more

Before the end of the year, they founded OpenAI.

Today, jury selection begins in Oakland — a nine-person jury asked to decide whether Altman and OpenAI president Greg Brockman get to keep their jobs. What follows is the private record made public: what the people at the top of the AI industry actually wrote to each other, in emails and texts and diary entries they never expected anyone to read. Every item below is sourced from court filings, depositions, and unsealed exhibits in Musk v. Altman et al., N.D. Cal. No. 4:24-cv-04722. And the reason all this stuff matters is that a jury will be asked to sort out not just the gossip — who betrayed whom, whether these were real partnerships or duplicitous back-stabbings — but whether OpenAI, which could IPO later this year, was founded on a lie.

Primary Court Documents

Before anything else, the documents themselves — the ones that made the judge decide this case had to go to a jury:

• Musk’s Amended Complaint — Northern District of California, Case No. 4:24-cv-04722. The full allegations, including the fraud and breach of fiduciary duty claims.

• Judge Gonzalez Rogers’ January 15, 2026 Summary Judgment Ruling — The 28-page ruling finding “ample evidence” and sending the case to trial. Cites Brockman’s diary directly.

• Unsealed Discovery Exhibits, January 2026 — 100+ documents including the Brockman diary entries, the Altman-Musk email chain, and the Zuckerberg-Musk texts.

Greg Brockman’s Diary

Greg Brockman, now president of OpenAI, has a journaling habit. During discovery in 2025, Musk’s attorneys subpoenaed his personal records. Three entries from 2017 — and I’m all for journaling, but careful, people — became the backbone of the judge’s decision to send the case to trial. Here’s some juicy stuff he somehow thought he should write down:

Read more

The only accountability mechanism that has actually gotten the people running the AI industry into a room, under oath — is a civil lawsuit, Musk v Altman. St…

Read more

On the night of January 2, 2026, a few hours before U.S. forces flew into Caracas and seized Venezuelan leader Nicolás Maduro, a soldier named Gannon Ken Van Dyke allegedly sat down and placed 13 bets on Polymarket. Van Dyke, a 38-year-old Master Sergeant with the U.S. Army Special Forces stationed at Fort Bragg, had spent the previous month helping plan and execute the operation. He had signed nondisclosure agreements. But he knew, in precise operational detail, what was about to happen. And he also seemingly knew he could make easy money on that information. According to prosecutors, he converted more than $32,000 of his personal savings into cryptocurrency, opened a new account under a pseudonym, and bet that Maduro would be “out” by January 31.

He won $409,881.

Federal prosecutors unsealed the indictment against Van Dyke on Thursday. He is charged with theft of classified government information, commodities fraud, wire fraud, and making an unlawful monetary transaction. He faces decades in prison.

Van Dyke's case marks the first time U.S. officials have leveled criminal charges against someone over prediction market wagers. The case is also the clearest illustration yet that in many cases (and especially in the most wildly profitable cases) these platforms are not forecasting tools at all, but transfer mechanisms — devices that move money from people who don’t know things to people who do. Now a new paper, drawing on the complete transaction record of a major prediction market, makes the most damning case yet against the industry.

The defense the industry has always offered goes like this: when money is on the line, people reveal what they actually believe, rather than what sounds good. That when people have financial skin in the game, it forces honesty, and therefore collective intelligence. Aggregate enough honest beliefs across a large and diverse crowd, and the market price converges on truth faster and more accurately than any expert, poll, or pundit. The CEOs of Polymarket and Kalshi have made this argument in congressional testimony, in press releases, and on television.

It has a real scientific lineage — the “wisdom of crowds” concept comes from decades of legitimate research showing that diverse groups, under the right conditions, produce better forecasts than individuals. This is why some of the most compelling and powerful findings in human behavior research involve setting people up around a table and experimenting on what they’re willing to bet real money on.

The research that launched the current public prediction markets was mostly conducted on closed corporate versions: employees forecasting their own company’s quarterly sales, or specialists estimating the probability of events in their domain. It was careful, bounded, and the participants were genuinely trying to aggregate dispersed knowledge.

Then Kalshi, Polymarket, and others made a business out of it. And what a business it is. Annual trading volume climbed from $15.8 billion in 2024 to about $63.5 billion in 2025 — a 4x surge in a single year. Weekly trading volume on Kalshi — which controls more than 90% of the U.S. prediction market — has surged to more than $3 billion today from about $100 million a year ago, according to Bank of America analyst Julie Hoover, who called Kalshi one of the “fastest growing non-AI companies” in the U.S. (The financial-analysis equivalent of calling something a miracle.)

Bernstein analyst Gautam Chhugani estimates total market volumes in 2026 will reach $240 billion — a 370% increase — and projects prediction market trading volume of $1 trillion a year by the start of the next decade.

Now a group of finance academics has tested what it is, exactly, that commercial prediction markets are delivering. Their paper — “Prediction Market Accuracy: Crowd Wisdom or Informed Minority?” — was posted this week. It is the most rigorous examination of this question yet conducted on a major live platform, using the full universe of transactions rather than a sample.

Accuracy on prediction markets, the researchers found, does not come from the crowd. It comes from roughly 3% of traders — a small group of informed participants whose trades predict final outcomes, move prices toward truth, and react to new information the moment it arrives. The other 97% of traders do not contribute to accuracy. They fund it. Their losses flow as profits to the informed minority.

The crowd isn’t getting smarter together. The crowd is getting fleeced by the people who already know the answer.

The industry has been using the wisdom-of-crowds argument as not just an intellectual selling point, but as a regulatory shield. Kalshi spent years in federal court fighting the Commodity Futures Trading Commission’s attempts to limit prediction markets, arguing that its election contracts served a public interest by aggregating information. And it won. The CFTC under the Biden administration concluded that election betting was contrary to the public interest; a federal court disagreed; the CFTC under the Trump administration dropped the appeal. The platforms have been largely unregulated since.

They have also been largely unregulated while generating a staggering volume of what appears to be insider trading. A Harvard paper published last month, drawing on public blockchain data, estimated that $143 million in profits have been earned on Polymarket by traders with insider information — covering more than 200,000 suspicious bets placed between February 2024 and February 2026. Van Dyke’s trading on what he knew is not an anomaly. It’s part of a pattern.

In the hours before U.S. forces struck Iran, an anonymous Polymarket account made roughly $550,000 betting that the strikes would happen and that Ayatollah Khamenei would be removed from power. In January, as Biden’s term expired, a single trader made $316,346 betting on five specific last-minute pardons — including Hunter Biden’s — even as the market odds of those pardons dropped toward zero. Columbia Law School’s Joshua Mitts, who advises the DOJ on insider trading cases, reviewed those trades and said the odds of the outcomes occurring by random chance were “virtually zero.” A trader known as AlphaRaccoon deposited $3 million into Polymarket and immediately bet across two dozen hyper-specific markets tied to Google’s Year in Search rankings — going 22 for 23 on bets, including a winning position on an obscure singer given 0.2% odds, shortly before Google accidentally published the results early and then pulled them. The account had previously won $150,000 predicting the exact release date of Google’s Gemini 3.0 before any public announcement existed.

Israeli authorities arrested several people in February, charging two with using classified military intelligence about the Iran strikes to trade on Polymarket. Van Dyke is the first American soldier charged, sure, but investigators said yesterday that similar suspicious bets were also placed around the Iran ceasefire negotiations in April, involving newly created accounts that reaped hundreds of thousands of dollars. (The White House reportedly warned staff against using confidential information to place trades the same day the Associated Press published those findings.)

Three congressional candidates were fined and suspended this week by Kalshi for betting on the outcomes of their own elections. The fines were $539.85, $784.20, and $6,229.30.

The industry’s response to all of this has been two-track. Publicly, the CEOs say insider trading is prohibited on their platforms. Privately — and sometimes not so privately — they have offered a different argument: that insider trading is actually a feature. Polymarket’s founder Shayne Coplan told 60 Minutes that people “going and having an edge to the market is a good thing” and that insider trading is “an inevitability” from which “benefits” arise. The logic is that insiders, by betting on what they know, move market prices toward truth faster than the crowd could without them. If an insider knows a military operation is happening tonight, and bets accordingly, the market price for that outcome rises — and thus, in theory, provides the public a more accurate probability estimate.

The new paper gives that argument partial scientific support: yes, the informed minority does improve accuracy. The 3% are genuinely pushing prices toward correct answers.

But here’s what that requires us to accept. The platforms are offering a product whose accuracy depends on a small group of people with access to information the public doesn’t have — classified military intelligence, corporate earnings data, the contents of presidential pardons before they’re announced — and whose profits come directly from the losses of the 97% who don’t have that access. The CEOs have been selling this as a tool for collective intelligence. The paper shows it is a mechanism for systematic extraction.

In the stock market, when insiders trade on non-public information for personal gain at the expense of other market participants, it’s a federal crime. The Securities and Exchange Act of 1934 established that principle, and courts have spent 90 years refining it. The specific reasoning is that the stock market functions on the assumption of fair play — that participants are competing on the quality of their public analysis, not on their proximity to secrets. When insiders trade on secrets, they are not contributing to a price-discovery process. They are exploiting an opaque wall between what they know and what you don’t for private gain.

Prediction markets have been allowed to develop without that framework in place. The CFTC, which regulates them, has insider trading on its list of enforcement priorities — as it announced in March. But it hasn’t yet brought a civil case. Van Dyke’s indictment is the first federal criminal charge. The instrument he used to commit the crime is still operating, still expanding, and still displaying its odds on Google Finance and in cable news chyrons as though they represent something the public collectively figured out. (CNN, where I’m a paid contributor, announced at the end of last year that it would partner with Kalshi and show its odds and predictions in its programming.)

If all of this screams “corruption” at you, try this on: Three bills are pending in Congress that would bar the President, Vice President, members of Congress, senior executive branch officials, and their families from trading prediction market contracts tied to political events. Donald Trump Jr., who is a paid strategic advisor to both Kalshi and Polymarket, and whose venture firm, 1789 Capital, invested tens of millions in Polymarket, has not commented on the legislation. His father, asked about Van Dyke yesterday, said he would “look into it,” wondered aloud whether the soldier had bet for or against the success of the operation, and mused that “the whole world unfortunately has become somewhat of a casino.”

There is a secondary problem that the industry’s accuracy argument obscures, and it concerns democracy more than markets. When prediction market odds are displayed on Google Finance and cited on CNN, they carry what researchers at the University of New South Wales have called “a veneer of objectivity” — they are backed by money rather than punditry, and money feels like evidence. Academic researchers have documented what they call “bandwagon effects”: as market odds shift toward a candidate, some voters shift toward that candidate, perceiving momentum and quality that the odds appear to confirm. The odds then move further. The market, which was supposed to reflect reality, starts to shape it.

In the 2024 presidential campaign, a single trader spent between $25 and $30 million on Trump-winning contracts. The researchers who documented this did not conclude definitively that it shifted the outcome. They did conclude that it shifted perception, and that the markets’ claimed role as neutral information aggregators collapses under those conditions.

It may be that some regulated version of prediction markets really can produce intelligence for the public worth betting on. The original Iowa Electronic Markets — the 1988 University of Iowa experiment that inspired everything Kalshi and Polymarket became — worked specifically because it was small, closed, and stripped of commercial incentive. The CFTC granted it a no-action letter on the condition that it operate solely for academic purposes, with no outside advertising, and a hard cap of $500 per trader. Within those constraints, it outperformed major polls in forecasting presidential vote shares roughly three-quarters of the time over sixteen years: true value to a democracy looking to better understand itself.

The design features that made the Iowa Electronic Markets work are precisely the features the commercial platforms have abandoned: a participant pool selected for genuine expertise rather than recruited for volume, position limits that prevent any single actor from moving the market, mandatory identity verification that makes insider trading traceable, and a narrow contract universe focused on questions where dispersed public knowledge actually exists. A regulated framework that restored those constraints — required identity, position caps, a prohibited-contracts list that extends to any market touching classified government information or the outcomes of events participants can directly influence, and genuine enforcement authority for the CFTC rather than the self-policing regime that let $143 million in suspected insider profits accumulate uncontested — might produce something genuinely useful.

The CFTC opened a public comment period on exactly this question in March 2026, with comments due April 30. The comment window closes in six days. (See the link in “Further Reading” to add yours.) What it produces will determine whether prediction markets become a civic tool or remain what the evidence currently suggests they are: a transfer mechanism dressed in the language of collective intelligence.

In the meantime, an information system that rewards insiders like Gannon Ken Van Dyke — who allegedly made 13 bets the night before the missiles fell, won $409,881, moved the money into a foreign cryptocurrency vault, and asked Polymarket to delete his account — misleads the public, and shapes the outcomes it claims to measure is not providing a public service. It is extracting value from the difference between what some people know and what most people don’t — and dressing the gap up as if it’s some sort of useful forecast.

Further Reading

• Gomez Cram, Guo, Jensen, and Kung — “Prediction Market Accuracy: Crowd Wisdom or Informed Minority?” (SSRN, April 20, 2026) — the peer-reviewed paper establishing that 3% of traders drive accuracy while 97% fund their profits

• NPR — “U.S. Soldier Charged with Insider Trading Over Maduro’s Ouster” (April 23, 2026) — primary reporting on the Van Dyke indictment

• Prediction Markets Grew 4X to $63.5B in 2025, But Risk Structural Strain (Yahoo Finance / CertiK) — annual volume figures from $15.8B in 2024 to $63.5B in 2025, with context on wash trading and sustainability concerns

• Prediction Markets in 2025: Data, Stats & Key Trends (Trade the Outcome) — full volume breakdown including the 127x growth since 2022 and category-level data showing tech and economics outpacing politics

• Prediction Markets Will Grow to $1 Trillion by 2030, Bernstein Estimates (CNBC) — Bernstein analyst Gautam Chhugani’s $240B 2026 projection and $1T by 2030 forecast, including Bank of America’s Kalshi volume figures

• How Prediction Markets Scaled to $21B in Monthly Volume in 2026 (TRM Labs) — transaction-level analysis of the growth from $1.2B/month in early 2025 to $20B+ by January 2026, including user wallet data and category shifts

• CNN — “Federal Prosecutors Exploring Whether Prediction Market Bets Trip Insider Trading Laws” (March 30, 2026) — DOJ/SDNY meeting with Polymarket

• NPR — “A Polymarket Trader Made $300,000 Betting on Biden’s Pardons” (April 16, 2026) — Bubblemaps analysis and Columbia Law sourcing on the Biden pardon trades

• Packin and Rabinovitz — “Prediction Markets as a Public Health Threat,” Science (April 2026) — peer-reviewed analysis of bandwagon effects and democratic manipulation

• Clinton and Huang — “Prediction Markets? The Accuracy and Efficiency of $2.4 Billion in the 2024 Presidential Election” — Vanderbilt study showing Polymarket called only 67% of 2024 markets better than chance

• CFTC Official Advisory on Prediction Markets (February 2026) — regulatory framework and documented cases

• Kalshi Becomes CNN's Official Prediction Market Partner After Raising $1B (Yahoo Finance / Cryptonews, December 2025) — the partnership that put Kalshi's real-time odds on CNN's air, including a live on-screen ticker overseen by CNN's chief data analyst; again, I am a paid contributor for CNN

• Three Economists Grabbed a Beer. A Multibillion-Dollar Industry Was Born. (NBC News, February 2026) — the founding of the Iowa Electronic Markets, its CFTC no-action letter and $500 position cap, and how its academic design constraints were shed as commercial platforms scaled

• CFTC Advance Notice of Proposed Rulemaking on Prediction Markets (Federal Register, March 16, 2026) — the formal regulatory comment process, open through April 30, 2026; the primary document for anyone seeking to shape what a federal framework actually looks like

• Prediction Markets at a Crossroads: Preemption, Enforcement and Rulemaking (Norton Rose Fulbright, April 2026) — the clearest current map of the federal-state jurisdictional fight, the CFTC’s enforcement priorities, and what formal rulemaking is likely to address

Other Currents

1. Meta logged its employees’ keystrokes, then announced their layoffs. The Model Capability Initiative — Meta’s new tool for capturing employee keystrokes and mouse clicks, framed as AI training data — was announced this week with no opt-out option. Two days later, 8,000 layoffs. The sequence is the story: document what the workers do, then eliminate them.

2. Palantir published a manifesto. Anthropic went to court. Two companies drew opposite lines in the same week. Palantir’s 22-point public document calls AI weapons inevitable, some cultures “dysfunctional and regressive,” and pluralism a failure. Bellingcat’s Eliot Higgins noted the obvious: these aren’t abstract ideas floating in space — they’re the stated ideology of a company that sells targeting software to militaries and immigration enforcement. Meanwhile, Anthropic has been in federal court since March fighting a Pentagon designation that labeled it a national security supply-chain risk — because it refused to remove safeguards against autonomous weapons and mass domestic surveillance. One company announced what it believes. The other is paying lawyers to defend it.

3. The Joint Chiefs said autonomous weapons are coming. Nobody asked whether they work. Yesterday the Chairman of the Joint Chiefs called autonomous weapons a “key and essential part of everything we do.” Anthropic’s actual position — that frontier AI is not reliable enough to kill people without a human in the loop — has not been rebutted. It has been bypassed.

4. 96,000 tech layoffs this year, all attributed to AI. Nobody has to prove it. Amazon, Meta, Microsoft, Snap, Salesforce, Block — every announcement uses the same language: efficiency, AI investment, right-sizing. No company is required to document whether AI replaced a single role. The NLRB has no jurisdiction over the framing. Nobody does.

5. Amazon’s Ring can now identify your neighbors by face. Familiar Faces rolls out facial recognition to Ring doorbells — identifying family, friends, delivery drivers — processed in Amazon’s cloud, opt-out by default. EFF and Senator Markey are pushing back. Three states have already blocked it. “Optional and disabled by default” is how every ambient surveillance feature starts.

6. OpenAI lost three senior executives in a single day and is projecting $14 billion in losses. The CPO, the head of Sora, and the enterprise CTO all departed the same Friday. Multiple cited the DOD contract and the cultural shift from research to commercial operations. The company generating $25 billion in annual revenue is simultaneously losing the people who built the products and spending $14 billion more than it earns.

7. Anthropic just passed OpenAI in revenue. The gap matters because of what each company agreed to. Anthropic hit $30 billion in annualized revenue this month, passing OpenAI’s $25 billion. Enterprise customers — not consumers — drove it. Worth noting alongside item 2: Anthropic is the company currently in court over autonomous weapons restrictions. OpenAI signed a DOD contract that removed equivalent restrictions weeks after Anthropic was blacklisted.

8. S&P 500 boards are disclosing AI risk while knowing almost nothing about AI. 83% of S&P 500 companies now list AI as a material risk in disclosures — up from 12% in 2023. Board directors with AI expertise: 2.7%. The people with the legal authority to set limits have mostly opted not to understand the thing they’re supposed to be limiting. This is the governance gap that makes every other story in this section possible.

Then came Ukraine. Russia deployed jamming systems that blacked out GPS signals across front-line areas stretching three hundred kilometers wide. Western precision bombs, the ones that need GPS to hit their targets, started missing. Ukrainian troops in some sectors navigated by pre-war paper maps. A February 2026 analysis in War on the Rocks concluded that any army operating today should simply assume GPS will be disrupted. The Government Accountability Office, Congress’s watchdog agency, has been tracking this problem for fifteen years. Its most recent major report, from 2024, found the military still hasn’t delivered a working jam-resistant GPS upgrade after more than two decades of trying and more than eight billion dollars spent.

The tool was too good not to use. And by the time the need to do without it arrived, the backup capability was gone.

The same dynamic is now visible in the military’s relationship with the AI company Anthropic — and this one is playing out in real time, in federal court.

Read more

Read more

Read more