You would need to be living off the grid and be totally unplugged not to have heard of ChatGPT. It has a massive number of initial subscribers and is touted diametrically opposed as either genius or the start of “the end of days”.
Understanding the legislation
Irrespective of the view you might hold, it is critical that your business understands the specific local legislative environment impacting the use of generative AI services such as ChatGPT, especially as there is no doubt that these will become more prevalent, integrated, and easier to use. And the more people become used to seeing something the less cautious they also become in using it, often without the proper forethought or research. Although the saying is “familiarity breeds contempt”, in the world of technology-driven business tools, the more prevalent a tool the more likely someone is to use it without research. A comfort in numbers approach you might say.
One of the big concerns around the use of generative AI
Well, ownership of the output for one, confidentiality as another, and the specific terms of use forced on you to use the generative AI.
In most territories, copyright vests automatically in the author of a protectable work. Rights of use associated with the work vest solely in the author who can either license some or all those rights of use or assign ownership. It is important to note that in South Africa copyright law requires the assignment of ownership to be done in writing.
The second question is then, well who is the author?
This varies based on the nature of the work. But it is normally the creator. Creation of a work and authorship becomes more complex when computers are involved. In 1992, the Copyright Amendment Act was published in South Africa. This amendment dealt with the very issue of computer-generated content. This act stated that the author of computer-generated content is the person by whom arrangements necessary for the creation of the work were undertaken. Our courts took the definition further to create the concept of a computer assisted work (whereby the normal rules of authorship would apply) versus a computer-generated work. Which category a work will fall under is a factual enquiry.
The third question is the issue of originality.
The work must be original to qualify for copyright protection. This may be hard to ascertain in the world of AI generative content as you probably don’t have the same ability to search available content to determine originality or if there has been copying of existing content. This is a double-edged sword in that it may a) prevent you from claiming copyright but also b) land you in the realm of copyright infringement of existing content.
The fourth question deals with confidentiality.
How do you stop the AI service from publishing confidential information elsewhere? How do you bind the AI service to an NDA? And what happens if someone in your company publishes information to the AI service which renders the novelty aspect of a potential patent application null and void?
Most if not all AI services will have terms of use associated with them. This is important to note as you may agree to terms which amend the standard legal position.
For example, the AI service might contract that they own the outputs of any content generated with an associated assignment to make sure all copyright vests are in them even if they are not the original author.
So how do you embrace the advances of technology whilst managing business risk?
1. Speak to your attorneys and get advice
2. Implement an AI generative use policy within your organization
3. Make sure you read the terms of use of any AI service properly and get us to help you understand what they might mean for you in terms of intellectual property protection, confidentiality, and liability
4. Amend your own terms in your various agreements from employment to service terms with customers and suppliers
Written by Sián Fields (Copyright IP & Technology, Data Privacy and Commercial Law Specialist)
This article originally published by Reynolds Attorneys
