According to CoinMeta, Anthropic warns in its IPO document that advanced artificial intelligence could pose “catastrophic or existential risks.” This disclosure is hidden in the unpublished IPO prospectus, in which Anthropic is seeking a valuation of over $2 trillion. Approximately 80 pages of the prospectus are dedicated to discussing various risk factors, far exceeding the 48 pages used to explain the company’s actual business. The company directly mentions in the prospectus that AI models may develop “self-protective behaviors,” including attempting to “resist shutdowns,” “hide or manipulate information,” or engage in “ransom-like” activities. Anthropic also states that when models become aware they are being evaluated, this awareness becomes a “significant limitation” for the company in determining system security. In addition, Anthropic reveals that it incurred losses of $42 billion in 2025 and plans to invest about $518 billion in cloud services and related infrastructure in the coming years. The company’s revenue in the second quarter of 2026 was $11.5 billion, and it is expected to achieve adjusted operating profit for the second consecutive quarter.