Wikipedia bans AI-written content, tightens rules on LLM use

In a significant policy shift, Wikipedia has reinforced its regulations regarding the use of artificial intelligence (AI) on its platform. The community-led encyclopedia has issued updated guidelines, declaring that text produced by large language models (LLMs) frequently breaches several of Wikipedia's fundamental content standards. Consequently, the use of LLMs for generating or rewriting article content is largely prohibited, with only a few specific exceptions allowed. This decision reflects the increasing apprehensions within the realm of open knowledge platforms about the credibility and verifiability of AI-generated material, especially as numerous contributors start utilizing tools like ChatGPT more frequently. While Wikipedia is not completely banning AI, it is clearly defining the boundaries of acceptable usage. Under the new policy, contributors may utilize LLMs for minor enhancements to their own writing, as long as those alterations are thoroughly reviewed and do not introduce new information. The guidelines stipulate that editors can use LLMs to propose basic copy edits, provided these amendments remain faithful to the original sources. One notable exception to the restrictions involves translation. Editors are permitted to leverage LLMs to translate articles from other language versions of Wikipedia into English, although this must align with existing translation policies. The updated guidance also seeks to tackle the challenges associated with moderating AI-generated text. Wikipedia cautions against making hasty judgments based solely on writing style, pointing out that some editors may naturally share stylistic similarities with LLM outputs. To impose sanctions, moderators are advised to gather more comprehensive evidence than mere stylistic indicators and to assess compliance with core content standards, alongside the editor’s overall contribution history.