Months before a mass school shooting in British Columbia, OpenAI employees raised alarm about the shooter's violent chatbot prompts and debated calling police, but ultimately opted not to, according to a new report.