In the second of a series of articles looking at how Generative AI is impacting trade marks and designs, we explore the need for designers to conduct their own due diligence when using Generative AI applications. (For further information, and to view the other articles in the series, please visit our Generative AI hub).
Generative AI
Artificial Intelligence (AI) refers to a type of computer program designed to mimic the way in which the human brain thinks, and uses this to spot patterns in data. AI programmes have been around for a long time, and have historically been used as analytical tools for data-heavy applications. Provided with an input (i.e. something to look for), an AI program can spot a pattern of data, and the user can then use this to make a decision. For example, the AI might be able to spot cancerous cells on an X-ray to assist in patient diagnosis, spot atmospheric changes to help predict the weather, or identify trends in financial data to help inform investment strategies, etc.
Generative AI refers to a sub-category of AI programmes that are able to not only spot patterns in data sets, but to then assemble these patterns into an intelligible work product that can be understood by humans. The most famous example of a generative AI system is the chatbot ChatGPT, which rose to prominence in 2022. ChatGPT is a type of generative AI system referred to as a “Large Language Model”, which describes systems that are able to “chat” with their users. ChatGPT is seemingly able to understand the meanings of any questions asked of it, and is able to formulate a coherent and useful response.
Following the success of ChatGPT, there has been a rapid proliferation of generative AI tools, which are able to create all sorts of different outputs for the user. For example, generative AI systems are now available for creating images, music, films, all at the push of a button.
Applications of Generative AI in design industries
With such a wide array of AI tools available, there has also been a sharp uptake in the use of these tools in design-heavy industries, such as graphic design, marketing, and product design. In these industries, AI tools are able to accelerate the design process by generating new content faster than it has ever been possible to do so before.
For example, because of its conversational interface and its ability to pull together information from disparate sources, many designers have used ChatGPT for idea generation. In one famous news article from the summer, a pizza restaurant in Dubai generated an entirely new pizza recipe based upon ChatGPT’s suggestion for “the best pizza for Dubai”. Although, with suggested toppings of pepperoni, blueberries and breakfast cereal, it seems unlikely that any top chefs will be quaking in their toques.
However, more practical uses of generative AI have also emerged. For example, the application “Clothing GAN” is an application for generating new designs of clothing based upon certain desired attributes. Clothing GAN is a type of AI program called a “Generative Adversarial Network”, which is able to generate new content having certain specified characteristics based upon the probability of those characteristics occurring within a source data set, in this case articles of clothing. By adjusting various sliders each associated with a certain feature (sleeve length, style, structure etc.) the user is able to generate an entirely new article of clothing, without the need for sketching or prototyping.
Case Studies
Whilst the potential efficiencies provided by the use of AI are certainly attractive, designers using AI should be aware that all which glitters is not necessarily gold. In particular, whilst generative AI systems are extremely good at generating new content, AI systems typically do not check whether that content is available for commercial use. Accordingly, without conducting due diligence, there may be a risk that the content generated by an AI system could infringe the rights of a third party.
For example, when we asked the image generating system Canva to “Design a sleek and luxurious mobile phone”, it provided us with two images (below). The first design (left) feels like a cross between a modern phone and an old-fashioned Nokia, blending as it does a smooth and shiny exterior with classic push buttons. This is quite an unusual combination, and might well be protectable as a design in its own right.
However, the second image (right) looks much more conventional, and appears to share a number of features in common with other popular models of mobile phone. In particular, the dropped screen notch at the top of the phone (containing the earpiece speaker, and forward facing camera) bears more than a passing resemblance to the iPhone 10, which is protected by Apple’s registered community design 003877091-0001 below. A commercial release of a new phone based upon the second image might therefore be expected to be objected to by Apple on the basis of its rights in the iPhone 10.
In another example, the image generator Midjourney was asked to “create a simple drawing of two people holding up an oversized heart, in the style of Keith Haring on a white background”.
In this example, the AI system has been told to emulate the style of a particular artist. Following the user’s orders, the AI has found the works of Keith Haring, learned from them, and recreated this style in the production of a new image. Given the large number of similarities between Haring’s original work (left) and the AI’s emulation (right), it is quite possible that this amounts to copying, and potential copyright infringement. Nevertheless, this is not clear-cut as there are still some amount of originality that has crept in, for example in the use of facial features, fingers, and boots / shoes. However, in light of the clear pointer to use Haring’s work, a commercial product based upon the AI generated image might well be objected to by the Haring Foundation.
Whilst this remains an evolving area of the law, generally speaking, it would seem sensible for designers to formulate their input prompts with as much of their own creativity as possible. This is perhaps best demonstrated by the light hearted example below, in which we asked Canva to generate an image of a “superhero IP lawyer”.
Whilst there is much for the author to be flattered by in Canva’s imaginings, it is notable that the superhero in the image generated bears more than a passing similarity to the fictional character Superman, who is owned by and is the copyright of DC Comics. Unlike the Haring example, the prompt did not contain any pointer or instruction to the AI system to borrow from or emulate Superman in particular. It would have been entirely possible for the AI system to generate an image of a generic superhero, perhaps using different coloured clothing and a different symbol. The AI’s decision to generate an image that so closely resembles a recognisable character should therefore cause concern.
Conclusion
Generative AI is certainly a useful tool for designers. However, as its popularity continues to rise, designers should be aware that they must conduct their own due diligence to determine whether or not the content created is, in fact, available for commercial use. For example, many online search engines offer reverse image searches, which can be used to find similar images and therefore identify potentially relevant third party rights. Moreover, designers should be aware of the rights, both registered and unregistered, held by competitor businesses. It is advisable to routinely check for the existence of relevant rights using an IP search provider.
(1) https://www.haring.com/!/art-work/812
(2) https://www.midjourney.com/jobs/ac518c34-4a6d-444e-9659-d9214300cd73?index=0