A multi-university study has found that large language models from different companies often produce strikingly similar responses, even to creative prompts, due to design processes that favor safe, ...