It's a way for AI labs to discuss safety while misdirecting from more mundane but widespread harms such as spam.