Artificial intelligence banner. Concept of innovation technologies in life. Vector isometric illustration of network, circuit connection of chip with hologram brain and computer, house and car

Implications of Generative AI for Intellectual Property  

The rise in prominence of generative AI has brought about changes in the way the world operates in many respects, and continues to change the legal and professional landscape as its capabilities develop.  

The term generative AI refers to artificial intelligence models which can generate new content after learning to recognise patterns and structures within a vast collection of training data fed to them. The model then uses this knowledge to create new data such as text, code, and images (among other outputs), mimicking these patterns and structures it has learnt – think ChatGPT (text output) and DALL-E (image output).  

In this article, we explore some potential challenges that might arise in the interplay between generative AI and intellectual property law.  

AI Can Generate Content That Falls Under Someone Else’s IP Rights  

One issue that could arise with the use of generative AI is the possibility for a user to (either accidentally or maliciously) generate content which infringes upon the IP rights of others.  

For example, using an AI-powered image generator to create a logo for your business could result in the generation of a logo that would be considered substantially identical or deceptively similar to another trade mark. This could lead to you being unable to register your logo as a trade mark, diminish the value of your brand by way of indistinctive branding making your business nonmemorable to customers or clients, or could leave your business vulnerable to action by the trade mark owner to oppose or remove your mark from the register.  

Use of IP-Protected Material to Train AI 

This issue can also apply to other IP rights, such as copyright which protects original artistic, literary, or creative works. In Australia, pieces of writing, images, music, sound recordings, and moving images fall under this category of IP protection automatically upon creation – there is no requirement for the owner to apply for or register the right. There have been multiple accusations of copyrighted content being used to train AI without the copyright holder’s permission. This issue of unauthorised use has been of particular concern to authors, with the New York Times suing OpenAI and Microsoft over the inclusion of published news content in their AI chatbots’ training datasets without authorisation.  

Recently a bill was introduced in the US that would force AI companies to disclose if they have used work protected by copyright to build generative AI systems, by submitting a notice to the Register of Copyrights – the Generative AI Copyright Disclosure Act. The bill has garnered support from many individuals and organisations across the creative industries. It will be interesting to observe the development of this bill, and the influence it could have on similar legislation regarding this issue in different jurisdictions.  

Inventorship and Ownership Issues 

Then, in patents, there is the issue of who to list as the inventor in cases where AI has contributed to the invention being protected.  

Australian patent law states that the inventor must be a “natural person” (i.e. a human being). An AI system does not make the cut and cannot be listed on a patent application as the inventor, as affirmed by the Full Federal Court in Commissioner of Patents v Thaler [2022] FCAFC 62. This case relates to a patent filed by Thaler listing DABUS (short for “Device for the Autonomous Bootstrapping of Unified Sentience”, an AI system he created) as the sole inventor, which IP Australia did not accept. It was subsequently appealed through multiple Courts and ultimately decided by the Full Court of the Federal Court of Australia — after being denied special leave to appeal to the High Court of Australia — that DABUS could not be listed as an inventor, let alone the sole inventor of an invention. We discussed this case and the decision more thoroughly in our article The Future is Now – Dabus Update 

The Designs Act 2003 also designates a person who created the design (or a person they were in contract with, a person deriving the title by Will or Law, or the legal representative of the deceased person) as eligible to be the owner of that design.  

This conundrum has posed an issue for courts in other jurisdictions worldwide, as well. Notably, the DABUS patent case has led to similar conclusions in the US, Europe, the UK, and New Zealand. Read our article about the US decision here

Hallucination 

Another issue to do with Generative AI and IP is hallucination – an AI tool generating new information that is not accurate, instead of producing a non-answer. This tendency for hallucination has been addressed as a trade marks issue in the aforementioned 2023 New York Times lawsuit against OpenAI and Microsoft in the US. The NYT alleged that the companies’ AI bots have provided incorrect information and attributed it to the Times, even making up articles that never existed in the first place. They claim that this misattribution associates the source with misinformation and misleading statements, damaging their reputation for providing reviewed information and diluting the New York Times trade mark due to its use alongside this incorrect information.  

Patentability Problems 

Due to the nature of AI, as a field that relies heavily on computers, it can be difficult to obtain patent protection for AI inventions. Australian patent law does not always lend itself well to computer-implemented inventions, since inventions which merely use known functions of a computer to produce known results have not been patentable. Generally, computer-implemented inventions must provide a novel and inventive way for the computer to work rather than using its known functions to complete a task.  

There are developments being made on this issue in both Australia and the UK – for more detail, see our article on A Place for Patent Protection in Australia’s AI Ecosystem.  

Key Takeaways 

Artificial intelligence is an area that poses some potential challenges for intellectual property law, as it stands currently. There are many decisions being made and pieces of legislation being updated to accommodate for developments in AI and guide the ways in which it is used in everyday, business and financial, and innovation and IP settings. Future intellectual property legislation may make express provisions to handle artificial intelligence and its contributions to IP. It is essential that those in the field keep up to date with these developments and the ways they can affect your business. 

If you or your business deal in the artificial intelligence space and require advice on the implications relevant to your intellectual property concerns, contact MBIP via our book a meeting form or email mail@mbip.com.au and a patent or trade marks attorney will be in contact with you to discuss your options.