Between opt-out and openness: how “AI Preferences“ are supposed to regulate the use of AI