The essay arrived weeks before a listing expected to value Anthropic at $965 billion, and it read like a treaty proposal for the age of artificial intelligence. Anthropic published a document of roughly ten thousand words calling on governments to negotiate a global “AI non-proliferation treaty,” with limits on frontier model development and international verification attached. The timing was not lost on Wall Street.
The company’s argument is straightforward: the most powerful AI systems are being built by a handful of labs, the risks of losing control of those systems are existential, and the world should treat frontier AI the way it treats nuclear weapons. Capability thresholds, export-style controls and an inspection regime all appear in the proposal, along with a plea for governments to act before the technology outruns the rules.
Analysts who follow the company read the document as something more layered than a safety statement. The doomsday narrative supports the valuation, they said: if the most advanced models carry existential risk, then the labs that build them safely deserve a premium. The safety-advocate identity, in the same reading, positions Anthropic inside the room where the rules get written, so that whoever sets the standards is setting them around products Anthropic already makes.
There is a logic to that sequence, analysts said. The people calling for brakes are often the ones who want you to slow down while they keep running. Latecomers, in this telling, have not even figured out the field before they must obey rules written by the first movers. A treaty negotiated now, with Anthropic at the table, would institutionalize the lead it already holds.
The same day the essay appeared, Forbes published its annual list of America’s richest self-made women, and Anthropic’s president, Daniela Amodei, landed at number two with a fortune of $15.5 billion. A year earlier the list had her at $1.2 billion. The jump of nearly thirteen times, driven by the surge in Anthropic’s private valuation ahead of the IPO, put the company’s story in a single number: the same week it asked the world to slow AI down, its own paper value multiplied.
The IPO itself is a study in how much confidence the market is being asked to place in one company’s trajectory. A valuation near $965 billion puts Anthropic within reach of the trillion-dollar mark, a threshold that took Apple decades to cross. The roadshow pitch, according to people familiar with the discussions, rests on demand for Claude models among enterprises, subscription growth and a computing advantage that scales with capital. Investors are being asked to pay for frontier research that has not yet produced the revenue base of the companies it is being compared with.
Critics outside the company call the treaty proposal a form of regulatory capture dressed in safety language. Rivals that publish open models argue that limits on frontier development would entrench the largest labs at the expense of everyone else, and parts of the AI research community have said the same. Inside the safety world, the response is split: some researchers praise the substance, others note that a treaty proposal from a company about to raise tens of billions of dollars is hard to separate from its balance sheet.
None of this is new terrain for Anthropic. The company was founded in 2021 by former OpenAI executives who left over disagreements about the direction of the industry, and it has always sold safety as a product attribute as much as a principle. What is new is the scale of the financial stakes. At $965 billion, the valuation has moved beyond the reach of the private markets’ usual yardsticks, and the essay gives investors a framework for why a company asking to be regulated should also be worth a trillion dollars.
The practical questions about the treaty remain unresolved, people close to the discussions said. Which governments would sign, how verification would work, and whether the United States and China could agree on anything at the frontier of AI are open questions the proposal does not answer. Anthropic’s own argument acknowledges the difficulty, asking for a conversation rather than claiming a solution.
The reaction from regulators has been muted so far, people familiar with the discussions said. The essay was written for governments, but the audience it reached first was the investment community, and the questions from that audience have been about the valuation rather than the treaty. Whether the proposal moves policy will take months to answer; whether the valuation holds will take weeks.
For investors, the calculation is simpler. The listing is expected to test whether the market can hold a near-trillion-dollar valuation on a company whose revenue is a fraction of that number. The essay gives the bulls a story, the bears a target, and the regulators a text to argue about. That is a combination the company’s founders, whatever their motives, could hardly have designed better for the weeks before the IPO.
