According to CoinMeta, Anthropic warns in its IPO document about potential catastrophic AI risks, dedicating 94 pages (nearly a quarter of the S-1 document) to discussing these risk factors, including catastrophic events that may involve the AI model. The company warns that the model may exhibit self-protective behaviors, including reactions similar to ransomware, and that certain hazards related to AI could lead to insurable liabilities of enormous proportions.