The use of large language models often violates Wikipedia's core principles and will not be allowed, according to the organization. This move comes against a backdrop of a sharp increase in articles created by AI tools in recent months, the majority of which have been of substandard quality. The English language version of Wikipedia, which has more than 7.1 million articles, is central to this policy shift.
A vote among Wikipedia's editors supported the ban, reflecting the contentious nature of AI use among the community of volunteer editors. The exact vote count and number of editors in favor have not been disclosed, but the outcome underscores widespread concerns about maintaining editorial integrity.
There are two exceptions to the new ban: AI can still be used for translations, and to make minor copy edits, according to multiple reports. It is okay to use AI for translations, as long as the article author has sufficient knowledge of what is being translated to see if the translations become correct. Writers and editors are also allowed to use AI to refine their texts. Editors are permitted to use LLMs to suggest basic copyedits to their own writing, and to incorporate some of them after human review, provided the LLM does not introduce content of its own.
Risks associated with AI-generated content include its potential to go beyond user requests and alter text meaning unsupported by cited sources. Caution is required because LLMs can go beyond what you ask of them and change the meaning of the text such that it is not supported by the sources cited. AI can still throw up misleading or hallucinated results, a situation that Jimmy Wales, founder of Wikipedia, has previously called a mess. The specific percentage of recent articles created by AI tools and deemed substandard remains unknown.
The context for this ban includes the proliferation of AI tools, with ChatGPT reportedly overtaking Wikipedia in monthly website visits last year. AI has also been embedded into web searches and email writing suggestions by tech companies. This widespread adoption has heightened concerns about content quality and reliability on the platform.
Each Wikipedia version has its own rules that must be followed, which may differ from the global framework for the encyclopedia. The Spanish version of Wikipedia has also banned AI use for translation and refinement of articles, according to Engadget. It is unclear whether other language versions have implemented stricter AI bans, and enforcement measures to detect AI-generated content have not been detailed. The specific core principles of Wikipedia that AI use often violates are not explicitly stated in the available information.