Chat GPT (“CGPT”) is an extraordinary language processing tool trained on a staggering amount of text data capable of understanding and generating human-like responses in prose form. Using CGPT, a user can pose questions (input) and generate answers (output) in the form of articles, essays, manuscripts, etc.
The application of CGPT at once raises legal issues such as copyright infringement, indicative that while CGPT is innovative and revolutionary, it is not risk-free. An example of such legal risk is presented in the Getty Images case. In January 2023, Getty Images sued Stability AI Inc. in London for copyright infringement, claiming that Stability AI had copied over 12 million images from its database to train its image generation model without permission.
In this edition of the IP Watch, we explore the issues around the use and application of CGPT in relation to copyright protection and infringement.
Chat GPT and Copyright
Copyright protection is given to the creator of a work to exclude others from substantial copying of the work. However, CGPT is trained/programmed to generate output using a collection of existing work and content created by someone else. These works are protected by copyright owned by the creators of the content. So far, there is no precedent that speaks to ownership of intellectual property rights (“IPR”) of content created by artificial intelligence. As such, it is unclear who owns the copyright in the output of CGPT – the platform or the user.
Not all works are copyrightable. Copyright does not protect ideas or thoughts but rather, the expression of the idea, i.e., ideas put in a fixed form. Under Nigerian law, copyright is tied to fixation and originality. That is, protection is given to works fixed in a definite medium of expression and works in which sufficient effort has been expended on making the work to give it an original character.
Ownership of IPR in the Output of Chat GPT
It is clear that a CGPT user owns the IPR in the input made into the platform. The ambiguity lies in the output. Who owns the IPR in the output of CGPT?
In Nigeria, copyright protection is given to works that are original and presented in a fixed form. In a legal sense, originality refers to independent creation and is not necessarily new or novel.
There are several doctrines of originality in copyright, including sweat of the brow, a modicum of creativity, and merger doctrine. The concept of originality in Nigerian copyright is similar to the doctrine of the sweat of the brow. Under the sweat of the brow doctrine, the courts emphasize the efforts and labour that were put into creating the work rather than the work’s novelty. Therefore, a sufficient amount of judgment and intellectual creativity has been required to create the work. Similarly, under the Copyright Act Cap C28 LFN 2004, the effort required to produce a copyrightable work must not be negligible or common. This begs the question, does CGPT output meet the threshold of originality under Nigeria copyright law.
Can a Chat GPT Output Meet Copyright Standards?
Having explored the possible legal risks presented by CGPT, especially in the area of copyright infringement, it appears that certain ‘human’ probationary checks and balances should apply before using the output generated by CGPT.
Since OpenAI has assigned ownership of the output to the user, salient considerations to address before using the output are:
- Whether the output may be identical to the work of another author who can claim copyright in work.
- Whether a user should obtain additional permission from creators of the content used in a CGPT output.