Anthropic’s 335-page initial public offering document filed Monday was packed with revelations about the firm’s internal workings and new insights into its customer base—including warnings about AI safety and misuse, revenue and safety-related expenses for its top executives as Claude’s maker prepares to go public.