Anthropic, the company behind AI chatbot Claude, has warned that artificial intelligence could pose "catastrophic or existential risks to humanity" - a sentence it tucked neatly into the paperwork for its planned blockbuster US stock market listing, presumably right between the risk factors and the footnotes.

The filing revealed losses of more than $80.7bn (£60.5bn) in the first half of the year, alongside $112.4bn in cash or equivalents and revenues of $16.3bn over the same period. The number of shares to be released, their price range, and a date for the listing have not been disclosed, because why ruin a good mystery.

Anthropic's valuation when it goes public has been expected to exceed that of SpaceX, which this summer became the biggest-ever stock listing at an initial valuation of about $1.8 trillion. No pressure.

In the filing, Anthropic said concerns had been raised by policymakers, regulators and researchers - as well as by people within its own organisation - that AI systems posed a significant risk to humanity. It said advanced AI systems "could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponised, or allowed to operate with a high degree of autonomy." Investors are advised to read that sentence twice and then perhaps lie down.

Anthropic will be the first true AI company to go public, making it a test case for the future of the technology. It said it will list on the Nasdaq under the symbol ANT, which is at least reassuringly small.

Founded just five years ago by chief executive Dario Amodei, his sister Daniela Amodei and five other co-founders, Anthropic announced its intention to go public in June. Its most significant rival, OpenAI, quickly followed suit, saying it, too, intended to go public - that listing is now expected early next year, according to a person familiar with OpenAI's plans. Nothing says confidence like a race to the exit.

After the companies decided on such plans, a wave of cyber-security and other technological breaches occurred involving their AI agents, or AI bots designed to operate somewhat autonomously. Current and former staff of both companies also gained public attention with new warnings that AI technology was being developed without the ability to protect the whole of humanity from possible future harms enacted by AI tools. The timing was described by observers as "not ideal."

In Anthropic's new documents, it reiterated promises to develop AI in a way beneficial to people, while simultaneously warning investors of its dangers. In its financial filings, Anthropic detailed losses that exceeded its revenue for the first half of the year, but said it would often make forward investments "that may seem irresponsibly large at first glance." At second glance, apparently, they look fine.

"In general, we will aim to be bold but sane, and to win trust with a track record of successful bets," the tech giant said, setting a low bar and then politely stepping over it.

Its success depends largely on its ability to develop the most advanced AI models, Anthropic said, adding that "any failure to do so could adversely affect our business" - a disclosure that ranks somewhere between "water is wet" and "the sun rises in the east" on the scale of corporate revelations.

Tech companies have been heavily investing in costly AI infrastructure and tools, like data centres and the informational models that underly all AI technology, pouring more than $1tn into the efforts over the last year alone. Because nothing says "we're being careful" like a trillion-dollar sprint.