Few companies open a sales pitch by explaining how their product might destroy the customer. Anthropic did exactly that this week, when it published the filing for its planned stock market debut. The document warns that advanced AI systems could pose "catastrophic or existential risks to humanity" if they are misused, replicated, weaponised or allowed to act with a high degree of autonomy. The financial figures are just as striking.

In the first half of the year, Anthropic reported revenue of $16.3bn, but its losses exceeded $80.7bn, roughly five dollars lost for every dollar earned. Even so, the company is expected to be valued above SpaceX, whose listing this summer, at about $1.

8 trillion, was the largest in history. Anthropic has not yet said how many shares it will sell, at what price, or when. It will trade on the Nasdaq under the symbol ANT, becoming the first fully AI-focused company to go public. The timing is delicate.

OpenAI, Anthropic's main rival, announced its own listing plans soon after Anthropic's June announcement, and it is now expected to float early next year, according to a person familiar with its plans. Since then, a series of cyber-security and other breaches involving AI agents, bots designed to work with some independence, has raised new doubts. Current and former employees at both firms have also warned that safeguards are lagging behind the technology.