The issue of ownership of intellectual property created by artificial intelligence (AI) is provoking heated discussions as the new chatbot “ChatGPT” takes the world by storm. It has raised questions about copyright in the work generated, ownership of copyright and even possible plagiarism of works by AI.
In 2021, Patent Offices around the world were taken aback by a patent application filed for an invention autonomously generated by an artificial intelligence, referred to as DABUS (DABUS is an acronym for “Device for the Autonomous Bootstrapping of Unified Sentience”). How is an invention created by a machine? Who owns the invention?
Although most patent offices have decided that AI systems cannot be cited as inventors, these applications have raised questions regarding the future of AI and the evolution of intellectual property law that is required to navigate ownership of these inventions.
In some ways this was reminiscent of the furore surrounding the ownership of copyright in the case of the “monkey selfies.” The case involved a macaque monkey in Indonesia who took a series of selfies with a photographer’s camera. While some claimed that the monkey was the rightful owner of the copyright in the photographs, the Ninth Circuit Court of Appeals in the United States ruled that the monkey could not be considered the copyright owner as copyright law only extends to works produced by human authors.
Given that the issue centres around the content ChatGPT generates, we couldn’t resist asking ChatGPT who owns the copyright in works created by ChatGPT.
Spoor & fisher: “Who owns the copyright in the work produced by ChatGPT?”
Chat GPT: “As an AI language model developed by OpenAI, the copyright in the works produced by ChatGPT would be owned by OpenAI. However, it’s worth noting that the output generated by ChatGPT is not intended to be copyrightable, as it is generated using artificial intelligence and does not necessarily reflect the creative expression of an individual human author.”
The fact remains – rights can only be assigned if they exist in the first place, and therein lies the rub. It could be argued that OpenAI own the copyright in the works because it is the software that generates the work based on complex algorithms that have been trained on vast amounts of data. It could also be argued that the human user who engages with ChatGPT to generate the works could own the copyright in the works because the human user is the one who creates the work through their interactions with ChatGPT, just as a writer or draughtsman creates a work using a word processor or drawing software. However, the ChatGPT output does not merely result from the human user’s inputs but also from complex algorithms that have been trained on vast amounts of data.
Another interesting argument is that no copyright work is created when content is generated using a language model like ChatGPT, because the resulting work is not the product of human creativity but rather the result of algorithms processing data.
For copyright to subsist in terms of our law, the work in question must be original. This doesn’t mean that it must be novel, have any technical or artistic merit, or be unique – it simply means that it must result from the author’s labours and should not be copied from another source. Our copyright legislation envisages that the author of a work could be a human being or a juristic person, so theoretically, OpenAI Inc. could be deemed to be the author or, at least, a co-author of a resulting work because it expended significant effort in developing the AI that enables ChatGPT to create the works.
It is also worth noting that the works generated by ChatGPT could be derived from other original works such as compilations of data used to train the language model. Some believe that the owner of the data should also own the derived work. But if a human being who writes a paper based on conclusions drawn from their investigation of all available information owns the copyright in the paperwhy, then, should the ownership of copyright in a work generated by ChatGPT from its assessment (albeit artificial) of available data be affected by the ownership of rights in that data?
Furthermore, if a resulting work is considered to be a derivative work, ChatGPT has been trained using a vast amount of data from a wide range of sources, so it would be near impossible to determine specifically which data contributed to the derivative work and, therefore, who owns the rights in the data and therefore, in the derivative work.
Put simply, the ownership of intellectual property generated by anything other than a human being is a complex question that our law may not be fully ready to answer! At this stage, it is unclear whether copyright subsists in works generated by non-humans, which is undoubtedly a premise that must first be evident before the question of ownership is tackled. Uncertainty is the only certainty until intellectual property law evolves to address these issues.
We had to clarify one last issue:
Spoor & Fisher: “ChatGPT, are you a lawyer?”
ChatGPT: “No, I am not a lawyer. I am a language model created by OpenAI that has been trained on a large corpus of text, including information about various topics, including law. However, my responses are based on the information that I have been trained on and should not be taken as legal advice. If you have a specific legal question, I would recommend consulting a licensed attorney”
As IP lawyers, we will continue to monitor development in this exciting field of law. According to ChatGPT our profession is safe for now. Still, the advent of this kind of technology is a flashing warning light to all of us – not only for schools and universities but also for legislators who will have to find a way to adapt to changing times and for legal service providers to take note of bots potentially dispensing legal advice.