An obituary test: AI's (probably unintentional) autocratic leanings in different languages
A simple experiment with AI-generated obituaries uncovers troubling biases in language models that show how they reinforce autocratic narratives across different languages.