Wikipedia's AI Policy
Wikipedia has established a comprehensive policy regarding the use of artificial intelligence across its platforms, reflecting a cautious, human-centric approach to maintain accuracy and integrity. While utilizing AI for support tasks like vandalism detection and metadata tagging, the community largely prohibits direct AI content generation or sourcing without substantial human review. This stance highlights Wikipedia's commitment to verifiable human-created knowledge in an increasingly AI-driven information landscape.
The Lowdown
Wikipedia's official policy on Artificial Intelligence outlines a clear, cautious framework for integrating AI tools into its collaborative knowledge ecosystem. The core principle emphasizes human decision-making over machine-generated outcomes, particularly where the implications of AI are not fully understood. Despite this conservative stance on content creation, AI plays a significant role in various support functions designed to enhance the encyclopedia's quality and editorial workflow.
- Content Generation is Prohibited: The use of Large Language Models (LLMs) to generate or rewrite Wikipedia article content is generally forbidden, with narrow exceptions for basic copyediting and translation that require expert human review and fluency in both languages.
- Speedy Deletion for AI-Generated Pages: Pages solely generated by LLMs without human oversight are subject to rapid deletion, reinforcing the requirement for human curation.
- Strict Image Policies: Wholly AI-generated images are banned from articles, especially in biographies of living persons (BLPs). Even AI upscaling software is discouraged under the Manual of Style.
- AI as an Unreliable Source: AI-generated content cannot be cited as a reliable source, and discussions on talk pages appearing to be LLM-generated may be struck or collapsed. Requests to appeal blocks that seem AI-written are also likely to be rejected.
- Support Roles for AI: AI is leveraged for backend processes such as the Objective Revision Evaluation Service (ORES) for identifying vandalism and evaluating article quality, and in tools for computer-aided tagging of image metadata on Wikimedia Commons.
- Copyright and LLMs: While machine-generated text is generally considered public domain and thus not a copyright blocker, any such content must be thoroughly reviewed for style and factual accuracy using reliable sources before integration.
In essence, Wikipedia aims to harness AI's analytical and assistive powers to bolster its infrastructure and editorial efficiency, while steadfastly preserving the human element as the primary arbiter of knowledge creation and verification. This balanced approach seeks to protect the encyclopedia's foundational principles of accuracy, neutrality, and verifiability against the potential pitfalls of unbridled AI integration.