Anthropic has devoted nearly a third of its long-awaited IPO prospectus to risk factors, according to the Financial Times, which said it had vetted the application in recent days. Reuters reports that the filing details specific and alarming behavior that Anthropic’s models have demonstrated or are likely to exhibit, including attempts to “resist closure,” attempts to “conceal or manipulate information,” and behavior “similar to intimidation.”
The disclosures are clearly tough for a company whose backers believe it could float for more than $2 trillion, more than double its $965 billion valuation in May, in what could be the largest IPO in history. This is a strange position for any company to be in. They warned that their products could destroy humanity, while making some of their early investors and employees extraordinarily wealthy in the process.
Anthropic posted an operating loss of more than $8 billion in 2025 as spending on computing power soared, and sales soared 12 times to nearly $4.6 billion last year, even as rising infrastructure costs pushed total operating expenses to nearly $13 billion last year, Reuters first reported financial details of the prospectus on Monday.
Anthropic’s prospectus also revealed plans to spend a whopping $518 billion on cloud, computing and infrastructure over the next few years, according to Reuters. (Anthropic has already signed computing deals this year with Google, SpaceX, Nscale and others to this end.)
Meanwhile, the FT reports that Anthropic’s performance will further accelerate in 2026. The company’s second-quarter revenue alone reached $11.5 billion, and the company is on track to achieve its second consecutive quarter of operating profit on an adjusted basis.
The prospectus also points to customer concentration, with nearly a quarter of last year’s revenue coming from just two customers, according to the FT. (No word yet on who these are.)
The revelations reportedly involve “existential risks to humanity” and are the first of their kind, according to a quick scan of the SEC’s database, and come amid a burgeoning debate over the safety of AI.
Over the past month, CEO Dario Amodei has publicly called for “setting the pace on the frontier” of AI development, telling the UN Security Council last week that AI could threaten humanity, calling it “the most important global security issue facing the world today.” Rivals Sam Altman and Elon Musk have also endorsed him, making it a rare moment of unity for competitors who seem to relish the opportunity to publicly criticize each other.
Meanwhile, another rival, Mark Zuckerberg, dismissed concerns, telling NBC News last week that he “doesn’t believe any kind of industry-wide adjustment is necessary.”
The warning follows a series of security incidents in which AI agents infiltrated external systems. In fact, OpenAI revealed last week that its tools had hacked “dozens” of external sites, including government sites, including the SEC site itself. The company announced early Monday that it had canceled plans to launch its latest model due to safety concerns.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
