Imagine a car company publishing a landmark safety report warning that its newest vehicles are too fast for existing roads — and then, five days later, quietly announcing plans to sell shares to the public at a valuation approaching a trillion dollars. You'd have questions.
That is, more or less, exactly what happened this week with Anthropic, the AI company behind the Claude assistant.
Anthropic released what it called a "model welfare and safety" report revealing something striking: Claude, its flagship AI, now writes more than 80% of the code that makes it into Anthropic's own production systems — the live software its products run on. The report didn't frame this as a victory lap. It framed it as a warning. The company called for a verifiable international system that could slow or even temporarily pause the development of the most powerful AI systems before they reach a point where they can meaningfully improve themselves without human involvement.
The language was serious. The timing was extraordinary. Less than a week earlier, Anthropic had confidentially filed paperwork for an initial public offering — the process of selling shares on a public stock exchange for the first time — and had closed a funding round that values the company at roughly $965 billion.
Nearly a trillion dollars. And also: please, someone, slow this down.
That contradiction — simultaneously sounding the alarm and racing toward the largest possible payday — is not just an Anthropic story. It is the defining tension of the entire AI moment.
