🤚 The Open-Palm Indictment

In the span of a single week, the American AI industry received a cease-and-desist from its own government, discovered that China had been running the largest intellectual property heist in AI history, and watched its most anticipated IPO get quietly postponed because one man refuses to be valued at anything less than one trillion dollars. If this were a screenplay, the studio would send it back for being “too on-the-nose.”

Let’s begin with the unprecedented. The Trump administration asked OpenAI to restrict the release of GPT-5.6 — the company’s newest model lineup consisting of Sol (flagship), Terra (balanced), and Luna (fast and cheap). This marks the first time the U.S. government has preemptively asked an American AI company to limit a model’s release before it launched. The White House’s Office of the National Cyber Director and Office of Science and Technology Policy cited concerns over “advanced cybersecurity capabilities” that could pose “unprecedented safety risks.” OpenAI complied, limiting access to “a small group of trusted partners,” while publicly noting that this arrangement “shouldn’t be the norm.”

Meanwhile, Anthropic told the U.S. Senate that operators affiliated with Alibaba and its AI lab ran the largest known AI distillation campaign in history. The numbers: 28.8 million interactions with Claude across roughly 25,000 fraudulent accounts, conducted between April 22 and June 5, 2026. The goal was to systematically extract Claude’s capabilities and repackage them without incurring the training costs. Anthropic’s Head of Policy, Sarah Heck, described the operation as conducted “illicitly, systematically, and at industrial scale.”

And then there’s Sam Altman, who looked at a potential 2026 IPO, looked at the valuation his bankers could guarantee, and said — we’re paraphrasing — “Not for less than a trillion.” OpenAI is now eyeing 2027.

👐 The Two-Handed Entanglement

These three stories are not separate developments. They are the same story, wearing three different hats.

The GPT-5.6 restriction reveals a government that is, for the first time, treating frontier AI models the way it treats advanced weapons systems: you don’t just ship them. The fact that all three models in the lineup — Sol, Terra, and Luna — were restricted suggests that even the “lower-cost option” crossed a capability threshold that made Washington nervous. OpenAI’s public pushback was notable: “We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.” Translation: we’ll comply this time, but don’t get comfortable.

The Alibaba distillation campaign is both breathtaking in scale and entirely predictable. This dwarfs the combined efforts of DeepSeek, Moonshot AI, and MiniMax, which Anthropic named in February for collectively generating more than 16 million Claude interactions. Alibaba nearly doubled that — alone. The technique itself — sending carefully designed prompts to a target model and training a cheaper model on the responses — is well-known. What’s new is the industrial scale: 25,000 fake accounts operating for six weeks, generating almost 29 million exchanges. This isn’t espionage. It’s a manufacturing process.

Two days after Anthropic’s letter to the Senate, the Commerce Department imposed strict restrictions on Anthropic’s own Fable 5 and Mythos 5 models, citing concerns that adversarial nations could weaponize their capabilities. The irony is exquisite: America restricts its own AI to prevent theft, while simultaneously discovering that the theft has already happened at a scale that makes the restrictions feel performative.

🌿 The Gentle Awakening

And then there is Sam Altman, standing in front of his bankers, being told that he can either go public now at under $1 trillion or wait until 2027 for a better shot at the number he wants. He chose to wait. SpaceX’s post-IPO wobble spooked OpenAI’s advisors — if Elon Musk’s rocket company couldn’t sustain its debut valuation, what hope does a company that still doesn’t turn a profit have?

But the real reason for the delay may be simpler than market timing. OpenAI confidentially filed its S-1 paperwork on June 8 and immediately said it “hadn’t decided on timing” because “there are things we want to do that are likely easier as a private company.” When a company tells you it prefers privacy right after filing to go public, what it’s really saying is: we have problems we’d rather solve without shareholders watching.

Prediction market Kalshi puts the most likely IPO announcement date before March 1, 2027. By then, Sonnet 5 will have been available for nine months, Alibaba will have presumably found new creative ways to steal things, and the U.S. government will have restricted at least two more models. The landscape Sam Altman IPOs into will not be the landscape he planned for.

👑 The Crown Verdict

What we are witnessing is the moment the American AI industry discovered that its three biggest challenges are, in fact, the same challenge: control. The government wants to control who gets frontier models. China wants to control — or rather, copy — the intelligence those models contain. And Sam Altman wants to control the narrative long enough to IPO at a valuation that starts with a “T.”

None of them will fully succeed. The government cannot restrict what has already been distilled. China cannot steal fast enough to keep pace with models that improve quarterly. And Altman cannot wait forever in a market where Anthropic just filed a $965 billion IPO of its own. The AI race is no longer about who builds the best model. It’s about who controls the moment — and the moment, as always, belongs to no one.

Inspired by US Government Blocks GPT-5.6, Alibaba’s AI Theft, and Why OpenAI Is Stalling Their IPO | MOONSHOTS by Peter Diamandis.

Your trillion-dollar filing is showing. Value wisely.