WikiHow Files Lawsuit Against OpenAI Over ChatGPT’s Use of Its How-To Library

“The complaint cited GPT-4 as stating that its ‘training data gave it enough exposure to wikiHow content to be able to generalize their visual patterns and formats effectively.’”

On August 21, wikiHow, Inc. filed a complaint in the U.S. District Court for the Southern District of New York against OpenAI, Inc. and eight affiliated entities, alleging violations of the Copyright Act and the Digital Millennium Copyright Act (DMCA). The lawsuit centers on claims that OpenAI copied wikiHow’s how-to articles without authorization to train ChatGPT and to ground its outputs, then used that copied content to generate substitutes for wikiHow’s website.

wikiHow described itself as a publisher of expert-verified instructional articles that has served billions of readers since its founding in 2005. The complaint stated that wikiHow hosts more than 500,000 articles and, at its peak, drew more than 150 million visitors a month. wikiHow owns 1,211 registered copyrights covering 11,211 articles at issue in the case, identified in an exhibit attached to the complaint.

According to the complaint, OpenAI has acknowledged training its large language models (LLMs) “on vast amounts of data from the internet written by humans.” wikiHow alleged that its articles were swept into that training data through crawling and scraping of its websites, whether directly or through intermediary datasets such as Common Crawl, WebText, and WebText2.