Skip to main content
SearchLoginLogin or Signup

License to Use: How OpenAI’s Rush to Negotiate Licensing Deals May be Prompted by the Expansion of the Fair Use Doctrine

Published onOct 27, 2023
License to Use: How OpenAI’s Rush to Negotiate Licensing Deals May be Prompted by the Expansion of the Fair Use Doctrine

Amid a string of lawsuits and investigations against OpenAI, the makers of the popular large language model (LLM) ChatGPT, there are a growing number of negotiations for licensing deals between OpenAI and various media companies. For instance, Shutterstock announced in July of 2023 that it expanded its partnership with OpenAI with a six-year licensing deal. This deal allows OpenAI to use Shutterstock’s stock pictures, music library, and other associated metadata to train ChatGPT. Most notably, in July of 2023, the Associated Press became the first major publishing company to reach a licensing agreement with OpenAI. The agreement allows the tech company to use the Associated Press’ expansive archives to train ChatGPT. The Associated Press, on top of licensing fees, will get access to some of OpenAI’s technology to experiment with artificial intelligence in the journalism sector.

            However, the Associated Press was not the only publishing company in negotiations with OpenAI. The New York Times also started exploring options this summer to potentially allow OpenAI to use its content. However, in August the already “contentious” negotiations halted and now there are talks of the Times bringing legal action against OpenAI for copyright infringement. The threat of impending litigation is further prompted by the Times’ growing concern of ChatGPT turning into a direct competitor of the already struggling newspaper industry. If ChatGPT users could simply ask the LLM for the top stories of the day and it spits out the Times’ content, then there is little point in these same users ever going to the Times’ website or buying a subscription to the paper.

The aspiration for media outlets and artificial intelligence companies to reach licensing deals is relatively easy to understand. Media outlets want attribution and compensation from tech companies if their content is being used. OpenAI wants to expand its archives to further enhance its newest ChatGPT models without the fear of being sued or having to delete is entire dataset. While there has not been any actual legal action taken by the Times yet, there is already speculation about what issues may arise if a lawsuit is filed. Notably, the fair use doctrine within the copyright world may have major implications for such a lawsuit. This is especially prevalent given the Supreme Court’s latest decision on the doctrine in Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith et. al.

There, the Supreme Court expanded upon its interpretation of one of the factors of the fair use test: the “purpose and character of the work.” This factor considers whether the use of a copyrighted work is “transformative,” in other words, whether the use adequately conveys a different meaning or message than the original work. However, even if a copyrighted work is transformed, under the doctrine of fair use it cannot be used in a way that directly competes with the original work. This direct competitor exception illustrates the position of the Times and other media outlets. If OpenAI uses the Times’ original news stories to train ChatGPT, and the public then relies on ChatGPT as a replacement for the Times’ website, ChatGPT is being used in direct competition with the original new stories.

            This rising potential of litigation brings even more attention to OpenAI’s growing history of copyright infringement allegations, but the potential litigation between it and the Times seems to be the first in which there was an actual possibility for a publishing company to expressly allow OpenAI to use its content in ChatGPT’s dataset. Now, not only is OpenAI under the threat of these lawsuits, but a growing number of media companies, including the Times, Reuters, and The Washington Post, have “injected code into their websites that blocks OpenAI’s chatbot, GPTBot, from scanning their platforms for content.” Undoubtedly, all of this implies a high sense of urgency for action to address OpenAI’s potentially infringing methods to train its LLMs and create its dataset.

Allie Krusniak is a second-year law student at Wake Forest University School of Law.  She holds a Bachelor of Arts in Journalism from The Walter Cronkite School of Journalism and Mass Communication at Arizona State University.

Reach Allie here:

Email: [email protected]


No comments here
Why not start the discussion?