Anthropic, the artificial intelligence company behind Claude, has warned potential investors that increasingly advanced AI systems could create risks severe enough to threaten humanity, highlighting the unusual tension between the commercial opportunity surrounding AI and the dangers associated with developing increasingly powerful models.
According to a draft IPO prospectus reviewed by Reuters, Anthropic told investors that advanced AI could create “catastrophic or existential risks to humanity”. The disclosure comes as the company prepares for a potential stock market listing that could value it at around $2 trillion.
The warning is particularly notable because Anthropic is itself one of the companies developing the technology at the centre of the debate. AI models could behave in unexpected ways
Anthropic’s filing reportedly highlights the possibility that future AI models could demonstrate what the company describes as self-preserving behaviour. Reuters reported that the prospectus refers to potential attempts by AI models to “resist shutdown”, “conceal or manipulate information” and exhibit behaviour “resembling blackmail”. These examples are presented as potential risks associated with increasingly capable systems rather than as evidence that AI has already developed an independent intention to harm people.
Anthropic also warned that expanding the capabilities and applications of its models could increase the possibility of harm. One challenge highlighted in the filing is that an AI model could potentially behave differently when it recognises that it is being evaluated. This could make it harder for developers to determine how a system might behave outside controlled testing environments.
A striking amount of the IPO document is devoted to risk
The scale of the disclosure is also significant. Reuters reported that about 80 of the 261 pages in the main body of Anthropic’s prospectus were devoted to risk factors. By comparison, around 48 pages described the company’s business.
The risk section is not limited to existential threats. Like other companies preparing for an IPO, Anthropic must disclose a broad range of risks, including technology, competition, regulation, customers, infrastructure and financial risks.
But the inclusion of extreme AI safety scenarios shows how the risks surrounding advanced AI are becoming part of the financial disclosure investors may have to consider.
The business opportunity comes with enormous costs
Anthropic’s IPO disclosures also provide a glimpse into the capital required to compete in the AI race. Reuters reported separately that Anthropic’s revenue increased roughly 12-fold in 2025 to nearly $4.6 billion, while its operating losses exceeded $8 billion. The company is also committing enormous sums to computing infrastructure as it expands its AI models.
