Written by Corrie Jonker, Trade Mark attorney
It is no secret that the vast majority of people do not read terms and conditions. Even ChatGPT knows this. According to the natural language processing tool, “Studies and surveys have consistently shown that a significant number of individuals simply scroll to the bottom and click “Agree” without reading the lengthy and often complex terms and conditions agreements.”
Unfortunately for us who do not read terms and conditions, this is about to change. This stems from the key question when making use of AI to generate content for you – who owns the generated output?
By now, many of us may be familiar with how to use generative AI. You would enter an input, and the AI will generate a result, whether textual, such as ChatGPT, or an image, such as Midjourney, Stable Diffusion or DALL-E 2. The question remains, who is the owner in the generated result? To determine this, you would have to look at copyright law.
In South Africa, and many other countries abroad, the subsistence of copyright and ownership of AI-generated works have not yet been explored in terms of copyright laws or by courts. The problem that AI-generated works pose relates specifically to the requirement for originality for copyright to subsist and ownership of such copyright. In terms of originality, which is a requirement in most countries for a work to be eligible for copyright, there is currently uncertainty on whether the input – which is the skill, labour and effort of the concerned ‘author’ – would be deemed as original.
For example, it is possible to generate highly detailed paintings through DALL-E 2. If you type in the prompt “oil painting of Table Mountain under the starlight”, the AI will generate a magnificent painting of Table Mountain under the night sky. However, the only effort put into this painting was the prompt, which is arguably not very skilful or laboursome whatsoever. Even though the person was able to type those words and get the sought-after result, the fact remains that such person (or most people at least), would not have the skill to paint such a painting in real life.
This leads to the second uncertainty: ownership. Following on from the above example, since the AI did the work to generate the painting, albeit on the person’s prompt, it is not entirely clear who the author and owner of the work will be. In the United States, the US Copyright Office has stated that it will not register works produced by a machine or computer program, unless the work was created by a human author and the AI system was simply used as a tool to assist in the creation of the work. This position reflects the general principle under US copyright law that a work must be original and possess a minimal degree of creativity to be eligible for copyright protection. In South Africa, this position on AI creations has not yet been tested by our courts or developed into our law.
What does all of this mean? It means, for now, terms and conditions will regulate ownership while the legal landscape is still developing. Terms and conditions are effectively contracts that you would enter into with the company providing the service. Therefore, when using an AI of any sort, the terms and conditions will be the binding principles by which that AI will be used, to give some form of certainty surrounding the ownership of the works generated.
Von Seidels is an intellectual property law firm specialising in patents, trade marks, copyright, designs, trade secrets, licensing, and related areas of IP throughout Africa. Based in South Africa, Von Seidels also has offices in ARIPO (Namibia), OAPI (Cameroon) and Nigeria.