Today, AI tools can generate new audio, text, video, and image content with just a textual input from users. The Harvard Business Review has shared that AI presents a transformative opportunity to “augment human creativity” and that these tools have the ability to “democratize innovation” by deeply involving users of products and services in the innovation process. Beyond their potential to aid innovation and “assist humans in their efforts to create hitherto unimaginable solutions,” AI tools have become popular for personal uses, such as when users flooded their Instagrams with AI generated selfies by Lensa AI in 2022. It is difficult to argue with the social, creative, and economic utility that exists within generative AI tools, but it is irrefutable that these programs have created a gray area in IP law, and more specifically, within copyright.

Generative AI models learn from mass datasets in order to produce content that is relevant to the AI tools’ objective (whether that be to create textual, visual, or audio-related content), and this data often includes copyrighted materials. This kind of learning and output model has caused issues across creative industries, most notably when earlier in 2023 a user called “ghostwriter” uploaded an AI generated song (“Heart on My Sleeve”) using voices that mimicked the artists Drake and The Weeknd. The AI generated song went viral on TikTok and was even submitted for a Grammy (but was subsequently denied). Universal Music Group, the record label home to both superstars, flagged that this content raises intellectual property concerns. The legality of generative AI is being challenged in other areas of the entertainment world, too, with many authors joining together to bring lawsuits against OpenAI, the creator of AI tools ChatGPT and DALL-E, and Meta Platforms, formerly known as Facebook, Inc. 

In July of 2023, comedian Sarah Silverman and authors Richard Kadrey and Christopher Golden filed a class action copyright infringement lawsuit in San Francisco alleging that Meta and OpenAI used copyrighted material to train their chat bots. In a subsequent court filing, the authors stated that “OpenAI violated U.S. law by copying their works to train an artificial intelligence system that will ‘replace the very writings it copied.’” Later in September, a group of writers, including John Grisham and “Game of Thrones” author George R.R. Martin, filed a separate lawsuit against OpenAI in New York. The writers argue that OpenAI’s language model and its creations are “derivative works” of their copyrighted materials, therefore infringing the authors’ copyright. It is difficult to predict how the courts will respond to these novel issues in IP law, but the U.S. Patent and Trademark office has opined that the generative AI training process “will almost by definition involve the reproduction of entire works or substantial portions thereof.” OpenAI even acknowledges that its programs are trained on “large publicly available datasets that include copyrighted works” and that this training process includes making a copy of the data to be analyzed. 

Whether the AI training process infringes the copyright of works in Silverman and related cases may be a factual question for the people. Last month, “a U.S. district court ruled that a jury trial would be needed to determine whether it was fair use for an AI company to copy case summaries from Westlaw…” to train their AI program to quote portions of legal opinions in response to questions from a user. While it seems likely that any AI company would argue that their training processes are fair use and therefore do not infringe on any copyright, this is a fact-specific inquiry and may vary across cases and juries. 17 U.S.C. § 107 lays out the four factors that determine if copying constitutes fair use: 

  1. “The purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes; 
  2. The nature of the copyrighted work;
  3. The amount and substantiality of the portion used in relation to the copyrighted work as a whole; and 
  4. The effect of the use upon the potential market for or value of the copyrighted work.”

For example, OpenAI contends that the use of copyrighted works to train its programs should be considered fair use under the factors above. In response to factor one, OpenAI argues that its use of copyrighted works are “transformative” because the training process creates “a useful generative AI system.” The company also argues that the third factor supports its fair use because copies are only used to train the program and are not shared with the public. OpenAI cites The Authors Guild v. Google, Inc. for support, noting “the U.S. Court of Appeals for the Second Circuit held that Google’s copying of entire books to create a searchable database that displayed excerpts of those books constituted fair use.” Generative AI companies may run into issues defending themselves under the fourth factor, though, considering that works like “Heart on My Sleeve” are similar to and in competition with the original works of the artists they are mimicking.  

While we cannot yet appreciate the substantial impact that generative AI will have on IP law, it is evident that challenges already exist in balancing the rights of creatives, the need for innovation, and the relevant protections within copyright law.