WikiHow sues OpenAI over ChatGPT training

WikiHow has sued OpenAI in Manhattan federal court, accusing the company of using its instructional articles to train and support ChatGPT without permission.

According to Reuters and the complaint, WikiHow says OpenAI copied more than 11,000 registered how-to articles, including through web crawlers and intermediary datasets. The filing also alleges that ChatGPT can reproduce protected passages and that OpenAI continued accessing WikiHow pages after the publisher blocked OpenAI crawlers through robots.txt.

The case matters because it targets a different kind of publisher from the news organizations already fighting AI training cases. WikiHow argues that ChatGPT can substitute for its practical step-by-step articles, not merely learn from them in the background.

WikiHow is seeking damages and an injunction barring OpenAI from operating models trained on its copyrighted works without authorization. OpenAI has argued in other copyright disputes that training AI systems on publicly available web material is fair use, but this complaint adds another pressure point around licensing, retrieval systems and publisher traffic.