Lawsuits claim OpenAI identified a “credible and specific” threat months before the attack but failed to alert authorities, raising new questions about AI companies’ responsibility.