Weeks after ChatGPT’s late-2022 launch, the International Conference on Machine Learning (ICML), one of the field’s top venues, issued a Large Language Model policy for its 2023 conference. The core rule: “Papers that include text generated from a large-scale language model (LLM) such as ChatGPT are prohibited unless the produced text is presented as a part of the paper’s experimental analysis.” In other words, an LLM could be a subject of study, but not a ghostwriter of the paper itself.
The policy carved out an explicit exception for editing. Authors were still allowed to use LLMs “for light editing of their own text,” analogous to how “LLM tools are used similarly to automate grammar checks, word autocorrect, and other editing tools.” ICML acknowledged the rule was deliberately conservative, written to guard against issues such as plagiarism, and said it expected the policy “may evolve in future conferences.”
Other major venues took softer stances. NeurIPS 2023 told authors they were “welcome to use any tool they wish for writing the paper” but that they remained “responsible for the entire content of the paper, including all text and figures,” and had to “ensure that all text is correct and original.” NeurIPS also required that any LLM important to a paper’s methodology be disclosed. The contrast - ICML’s prohibition versus NeurIPS’s disclose-and-take-responsibility approach - framed the debate that ran through scientific publishing for the next several years.
Why business readers should care: the ICML and NeurIPS policies were among the first institutional attempts to govern generative AI in professional writing. The two models they established - prohibit generation versus require disclosure and human accountability - recur in nearly every later AI-use policy, from newsrooms to law firms.