Who Benefits, Who Gets Hurt, What Stops It

Anthropic lists safety-aware judgment as one of four core PM competencies, not a bonus round. OpenAI candidates get asked to reason through disparate impact before anyone says the word ‘ethics.’ The question underneath is always the same: who benefits from this system, who could be harmed by it, and what specifically stops the harm before it happens.

A safety mock interview is good practice. It’s still a drill, and it’s the version currently on the shelf. The stronger version of this story is a real decision, on a real project, where the safeguard changed what got built. That’s the harder story to tell, and the one that actually separates a candidate.