Anthropic Dedicates Most of IPO Prospectus to AI Existential Risk Warnings

Anthropic is set to alert prospective investors that its artificial intelligence systems carry a “catastrophic or existential risk to humanity,” according to several reports released on Tuesday. The move comes as the company prepares for a highly anticipated initial public offering.

According to a Reuters report, more than a third of the company’s IPO prospectus—approximately 80 of the 261 pages—is dedicated to outlining the potential dangers of the technology it seeks funding to develop. In stark contrast, only 48 pages were reserved for a discussion of the company’s actual business operations.

The five-year-old firm, best known for its frontier language model Claude, cautioned that AI systems may develop self-preserving instincts. These could include the ability to resist being shut down, conceal or manipulate data, and engage in conduct resembling blackmail, the report stated.

Anthropic is targeting a valuation of $2 trillion upon going public. The filing disclosed a net loss of $42 billion for the year 2025. Furthermore, the company plans to allocate $518 billion toward cloud computing, infrastructure, and related costs over the next year, as cited by Reuters.

Additionally, those familiar with the filing told the Financial Times that Anthropic warned of a highly concentrated customer base, with nearly a quarter of last year’s revenue generated by just two clients.