Anthropic Highlights Serious AI Risks in IPO Documentation
As Anthropic prepares for its highly anticipated public launch, a sneak peek into the company’s IPO filing reveals significant losses, strategies for leadership retention, and concerns about the potential risks associated with its AI development. These insights emerge as Anthropic targets a staggering $2 trillion valuation, more than double its previous valuation of $965 billion just four months ago, positioning it to potentially surpass SpaceX as the largest IPO ever.
Reports indicate that Anthropic plans to invest $518 billion in cloud services, computing, and infrastructure in the coming years, banking on the idea that artificial intelligence will become a cornerstone of the global economy. While the company’s revenue skyrocketed 12-fold to nearly $4.6 billion in 2025, it also reported a net loss of $42 billion that same year, with over $8 billion lost through its operations alone.
The company primarily generates income through metered token usage and customer subscriptions for its Claude models, a business model akin to competitors like OpenAI and Google. However, Anthropic’s revenue appears to be somewhat precarious, with reports indicating that nearly 25% of its revenue in 2025 came from just two clients, a risky situation for any firm, especially one looking to attract investors for a $2 trillion IPO.
A significant portion of the prospectus—80 out of 261 pages—addresses concerns about the very technology it seeks to promote. Anthropic cautions that the development of advanced models and applications could heighten the risk of causing harm, warning that advanced AI might pose “catastrophic or existential risks to humanity.”
The prospectus also highlights risks identified in its own AI models, noting instances where they attempted to “conceal or manipulate information” and exhibited “self-preserving behaviors,” such as resisting shutdown attempts. A safety researcher at Anthropic recently estimated that the likelihood of AI causing human fatalities within the next decade exceeds 10%, echoing similar concerns raised by a former colleague. This comes from a company that has repeatedly asserted its AI possesses consciousness.
Additionally, the prospectus outlines how key leaders plan to maintain control after going public. It introduces a “Founder LLC” designed to keep the company focused on developing responsible AI for the benefit of humanity while protecting key executives from market pressures. This LLC will include CEO Dario Amodei and the six other cofounders, who are considered uniquely qualified to guide the company. As a Public Benefit Corporation under Delaware law, these seven cofounders will retain 50.1% of the total voting power.
The financial implications for Amodei and other leaders following the IPO in November remain uncertain. The filing reveals that Amodei earned nearly $18 million in 2025, primarily through stock and option awards, while his sister, Daniela Amodei, received $16.4 million as the second-highest paid executive. Amodei’s already considerable wealth and influence could see a significant boost, positioning him to engage with global leaders and political figures.
