For its 2023 conference, NeurIPS - one of the largest machine-learning research venues - took a disclose-and-take-responsibility approach to large language models rather than banning them. Its call for papers told authors they were “welcome to use any tool they wish for writing the paper,” but that “authors are responsible for the entire content of the paper, including all text and figures,” and “must ensure that all text is correct and original.” The policy also required that any tool important to a paper’s methodology, including an LLM, be described in the paper. This stood in contrast to ICML 2023, which the same season prohibited LLM-generated text outright unless it was part of a paper’s experimental analysis.