By Sarah Simpson and Tegan Miller-McCormack

Love it or hate it, artificial intelligence, and particularly ChatGPT, has flooded the news in recent months. Whether you are a lover or a hater of artificial intelligence, the frenzy around ChatGPT has brought with it the reality that artificial intelligence is no longer for use by a select few but a tool sure to become more prevalent in our everyday lives. With that in mind, it is crucial to understand how artificial intelligence holds up in a virtual world within a legal framework developed to protect individuals' intellectual property and personal data in the physical realm.

ChatGPT and intellectual property

ChatGPT is an artificial intelligence chatbot developed by OpenAI, which has been trained and built on large language models (currently GPT-4), which allows users to ask questions or make requests of ChatGPT in return for human-like answers. GPT-4 (and other language models) uses data from books, websites and articles to provide an answer.

While there is widespread adoption of ChatGPT work product, important questions remain about information accuracy, ethics and societal connotations. For example, in academia, the International Baccalaureate (IB), a global leader in international education, gave the green light on using the software, accepting that artificial intelligence will be part of everyday life. The IB did so because it does not regard any work produced by these tools to belong to the students using it, and any use of the tools should be quoted and referenced. Such use does not raise intellectual property concerns under English law, as the use by students is private and non-commercial, meaning it is an exempt activity under the English Copyright, Designs and Patents Act 1988 (CDPA). However, it is interesting to note that the IB regards the author, and subsequently the first owner, of the work to be the chatbot rather than the student operating it.

Beyond academia, what are the implications for users of ChatGPT and other chatbots in commercial settings? There are copious TikTok hacks showing how to talk to ChatGPT to optimize its output for use at work. For example, marketing managers can ask ChatGPT to write a social media post taking into account engagement for their client. We have to ask, though, would the marketing manager be infringing copyright at any point by doing so? What if a music artist gets writer's block and asks ChatGPT to write a song, then the artist goes on to perform it for money? Who owns that copyright? Is copyright infringed at any point? If so, who infringes it?

Source material

ChatGPT relies on source material for the language model and training to provide its output. This source material may be subject to certain intellectual property rights and therefore be legally protected. However, if ChatGPT is not directly copying or reproducing the source material but is simply deriving inspiration or learning from such material, can it be alleged to infringe on the intellectual property held in the source material? Some generative AI companies, such as Stable Diffusion and Midjourney, are facing legal challenges in the US by artists claiming that the AI tools have been trained to use images without the consent of their owners; these tools create images from simple text instructions by analyzing existing images including illustrations, artwork, and photographs.

Following a consultation into AI and intellectual property, the UK Intellectual Property Office (UK IPO) proposed a new exception for copyright law in June 2022. Specifically, where text and data mining systems extract data and copy works for any purpose (without a license), it would not constitute copyright infringement (TDM exception). There would be no option to allow rights holders to opt out of this exception.

However, following significant backlash from creative industries, UK government representatives have publicly suggested that the proposed TDM exception will not proceed and, instead, further engagement with the creative industry is anticipated. The Design and Artists Copyright Society, which represents visual artists, has warned that "this change will have far-reaching detrimental consequences" and has urged the Government to "look again at how the policy objectives" of supporting AI-driven technologies "can be better met without undermining creators' rights." Given the government's intention to become a hub for digital innovation and to have an attractive regulatory framework for the industry to thrive, it is likely that we will see further developments in the law favoring AI technologies, which may be of concern to rights-holders who will want to ensure that their intellectual property is protected.

Ownership

Section 9(3) of the CDPA states, "in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken." This is supplemented by section 178 of the CDPA, which defines a computer-generated work as one that "is generated by computer in circumstances such that there is no human author of the work." The "person by whom the arrangements necessary for the creation of the work are undertaken," is still rather ambiguous and it is unclear whether the owner is the programmer, the user or both.

We, therefore, asked ChatGPT its opinion on this matter: "Who owns the copyright of content generated by ChatGPT and on what basis?" ChatGPT gave the following response: "The ownership of the copyright… would depend on the specific terms and conditions agreed upon by the user and OpenAI… OpenAI would likely retain the copyright to the output generated by ChatGPT as it is the owner of the technology and the data used to train the language model. However, the specific terms of use and license agreements could vary depending on the intended use of the output and the agreement between OpenAI and the user."

As the platform is currently configured, each user is subject to the general terms of use. The subject of ownership of works is narrowly addressed by OpenAI's sharing and publication policy. The policy discusses "content co-authored with the OpenAI API." It states that creators wishing to publish their written content created in part with the OpenAI API have permission to do so under certain conditions, such as that the "role of AI in formulating the content is clearly disclosed in a way that no reader could possibly miss."

This still raises questions about whether creating content such as marketing posts fall under this provision, and if so, does OpenAI regard this as co-authored? Moreover, if the output of ChatGPT is not edited at all, is this still co-authored? While it is likely that the intention of discussing co-authorship also implies co-ownership, the terms of use of ChatGPT are still unclear as to who owns what in legal terms. As the technology develops and the extent of ChatGPT's use in the ordinary course of business expands, OpenAI may release more extensive terms. Alternatively, perhaps the platform will implement a paywall with terms assigning ownership.

Infringement

Another matter to consider is if ChatGPT created an original work for one user (User 1) but then went on to generate the same work for another (User 2). Would User 1 be able to claim copyright infringement? Given that to establish copyright infringement, User 1 would need to prove copying and a causal link demonstrating a connection between the two works, it may be difficult for User 1 to prove infringement against User 2 if User 2 did not knowingly or unconsciously, copy their work. This still raises issues that will need to be ironed out either by the ChatGPT terms of use or by legislation, as copying of the work still may have ramifications for User 1 that would need to be addressed.

ChatGPT and data protection

ChatGPT raises various issues and questions with respect to intellectual property considerations and creates a host of concerns regarding the protection of personal data. While ChatGPT does not provide any personal data, there is a risk that it could, and there is the risk that it is still processing personal data in its training bank. In March 2023, Italy's data protection authority, Garante per la protezione dei dati personali (Garante), banned ChatGPT over data privacy concerns finding that the platform violated various articles of the GDPR, such as having no lawful basis for processing and no age verification measures. Garante has since informed ChatGPT that it can operate in Italy if certain conditions are met.

This month, the European Data Protection Board also launched a task force to investigate ChatGPT "to foster cooperation and to exchange information on possible enforcement actions conducted by data protection authorities." While countries in the EU are considering bans or EU-wide regulation, the UK Information Commissioner's Office put together "eight questions that developers and users need to ask" when it comes to generative AI:

  • What is your lawful basis for processing personal data?
  • Are you a controller, joint controller or processor?
  • Have you prepared a Data Protection Impact Assessment?
  • How will you ensure transparency?
  • How will you mitigate security risks?
  • How will you limit unnecessary processing?
  • How will you comply with individual rights requests?
  • Will you use generative AI to make solely automated decisions?

What's next?

Thousands of notable individuals, from Elon Musk to Steve Wozniak, signed an open letter to call on AI labs to immediately pause the training of AI systems more powerful than GPT-4 for at least six months. The letter states that "powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable."

Indeed, data privacy concerns and intellectual property issues raised are the tips of the iceberg. Technology is developing at an exponential rate and, at the same time, is creating a web of legal challenges. Generative AI is a space that needs regulation and codes of practice. While the government is focused on the need to make legislative changes, the pace at which those changes are implemented is often much slower than technological development. It is likely, therefore, that many of these issues will be put to the test and left to the courts to determine how generative AI fits within our current protections.

Hayley Rabet, a Katten Muchin Rosenman UK LLP trainee, contributed to this article.

To read the full newsletter, please click here.