What fatalism? OpenAI has already made security improvements and they will make more. I also think having common standards for containment might help.