Fair Use

The core dispute in AI copyright litigation is whether the large-scale ingestion of copyrighted works during the AI model training phase constitutes "fair use"; proponents cite the Google Books case as precedent, while opponents point out that AI models can sometimes reproduce training data nearly verbatim.

Fair use is a principle in U.S. copyright law that permits limited use of copyrighted works without the copyright holder's permission for purposes such as commentary, news reporting, teaching, and research.

The fair use doctrine established under Section 107 of the U.S. Copyright Act was codified in 1976, and courts must weigh four factors: the purpose and character of the use, the nature of the copyrighted work, the amount used, and the market impact.

A court's assessment of fair use must consider four factors: the purpose and character of the use, the nature of the copyrighted work, the amount of the original work used, and the effect on the market for the original work.

Fair use analysis under U.S. copyright law requires weighing four factors, with the transformativeness of the use and the impact on the original work's market being the key points of dispute.

There remains significant debate in academia over whether AI model training constitutes 'fair use'.

AI companies argue that model training constitutes 'transformative use,' but rights holders counter that the training process involves complete copying of original works, and that generative AI is directly competing with and substituting for the original works market.

Section 107 of U.S. copyright law permits unauthorized use of copyrighted works in contexts such as commentary, education, and research (the fair use doctrine), but whether commercial AI training qualifies as transformative use remains disputed.

The Fair Use doctrine established by Section 107 of the U.S. Copyright Act is the primary defense basis for AI companies, with the core argument being that model training constitutes transformative use.

合理使用是美国版权法第107条确立的例外原则，判断标准涵盖使用目的、作品性质、使用比例及对市场的影响四个维度

If courts rule that AI training constitutes copyright infringement, legitimate activities such as search engine web indexing and caching, text mining, and large-scale data analysis in academic research could all be implicated

OpenAI maintains that its model training falls within the scope of Fair Use.

The fair use doctrine can be traced back to the case law rule established by federal judge Joseph Story in the 1841 case Folsom v. Marsh, and was later codified in Section 107 of the U.S. Copyright Act in 1976.

There is no unified jurisprudence worldwide on the legal dispute over the boundaries of fair use in AI training, and multiple lawsuits in the United States are still being heard.

In the context of AI training, tech companies often argue that using web content for model training constitutes 'transformative use' and should fall within the scope of fair use.

Content providers believe that large-scale reproduction for commercial model training exceeds the boundaries of fair use.

As generated content increasingly competes with original works in the market and multiple lawsuits proceed, courts and legislators are beginning to re-examine the boundaries of the "fair use" defense.

In several lawsuits in recent years, courts have taken a cautious stance toward the practice of "bulk copying of complete works," especially when the training data includes entire academic books, and the four-factor analysis often works against defendants.

Multiple media outlets have sued AI companies for scraping content without permission to train models, with core disputes including the boundaries of fair use, the legality of scraping, and market substitution effects.

In jurisdictions such as the United States, AI companies often invoke "transformative use" to claim fair use, but this defense faces growing challenges in the news content domain.