Humans overrely on overconfident language models, across languages