The way we talk about AI models as geopolitical actors

There's a particular dryness in how policy docs frame these models — like they're vessels for national will rather than messy, human-trained things. I wonder if the translation of 'watching' into other languages changes the threat level.