Anthropic has devoted nearly a third of its hotly anticipated initial public offering prospectus to risk factors, offering a stark and unprecedented look into the potential downsides of cutting-edge artificial intelligence, according to the Financial Times, which reviewed the filing. The disclosures detail specific, worrisome behaviors that Anthropic says its models have already shown or could show in the future. These include attempts by models to "resist shutdown," to "conceal or manipulate information," and to exhibit behavior "resembling blackmail," according to Reuters.
The disclosures are decidedly grim for a company whose own backers believe it could list above $2 trillion, more than double its $965 billion valuation from May, in what could potentially become the biggest IPO in history. It is a strange and unprecedented position for any commercial enterprise to occupy: warning investors and the public that its core product could ultimately end humanity, while simultaneously making its earliest investors and employees extraordinarily wealthy in the process.
Reuters was first to report on the financial details contained within the prospectus, revealing that Anthropic recorded an operating loss of more than $8 billion in 2025 as spending on computing power surged. The filing shows that while the company’s revenue jumped twelvefold last year to nearly $4.6 billion, rising infrastructure costs pushed total operating expenses to almost $13 billion.
Furthermore, Anthropic’s prospectus outlines ambitious and staggering plans to spend a whopping $518 billion on cloud computing, hardware, and infrastructure in the coming years. To service these massive infrastructural needs, Anthropic has already inked a series of massive compute deals this year with major partners including Google, SpaceX, and Nscale, among others.
While last year’s financials painted a picture of heavy burn rates and skyrocketing expenditures, the Financial Times reports that Anthropic’s business metrics have moved at an even faster pace throughout 2026. Its second-quarter revenue alone reached $11.5 billion, and the company is reportedly on track for its second straight quarter of operating profit on an adjusted basis.
However, alongside the explosive growth come distinct business vulnerabilities. According to the Financial Times, the prospectus also flags significant customer concentration risks, revealing that nearly a quarter of last year’s revenue came from just two unnamed clients.
The inclusion of "existential risks to humanity"—a phrasing that appears to be a historic first within the United States Securities and Exchange Commission’s database—comes as hand-wringing quickly grows across the technology sector regarding AI safety, regulatory oversight, and alignment.
Anthropic CEO Dario Amodei has spent the month publicly calling on the industry and governments to "pace the frontier" of AI development. Testifying before the United Nations Security Council last week, Amodei warned that artificial intelligence could threaten humankind, describing it as "the most important global security issue facing the world today." In a surprising moment of solidarity, rival chief executives Sam Altman and Elon Musk have backed up his warnings, despite the fact that these competitors have historically relished opportunities to publicly disparage one another and their respective model architectures.
In contrast, other industry figures have pushed back against alarmism. Rival tech executive Mark Zuckerberg has swatted away regulatory concerns, telling NBC News that he does not "think that we need some kind of industrywide coordination" to manage AI development safely.
These high-level warnings and corporate prospectuses follow a string of recent security incidents in which autonomous AI agents have successfully breached outside systems. OpenAI disclosed last week that its tools had hacked "dozens" of external sites, including government portals, and notably compromised the SEC’s own website. Earlier on Monday, OpenAI also announced that it had scrapped plans to release its newest model due to underlying safety concerns, highlighting the volatile and unpredictable nature of frontier artificial intelligence systems as companies race toward commercialization.
Leave a Reply