Artificial Intelligence and IP Ownership: will the IP created by AI meet its maker?

Technology often develops quicker than the laws which are supposed to regulate its use. For this reason, the law is often reactive rather than proactive in dealing with technological advancements. The development of artificial intelligence (AI) and machine learning is no exception to this rule.

The law in relation to intellectual property (IP) is constantly evolving and being applied in new ways as the tools and technology used to create IP, and the medium of the IP itself, becomes more advanced. In a world where we are developing programs and machines which are designed to learn, think and problem solve with less and less human intervention, it raises a wide variety of legal and ethical questions. Perhaps one of the most interesting questions being, how does the law recognise IP created by AI and how does it address issues of IP ownership and infringement in that context.

What is AI and what are its uses?

AI is defined broadly and has many different definitions adopted by different computer scientists. The general consensus is that it relates to the engineering of machines and computers to replicate and understand human intelligence.

It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable”[1]

Whether a computer program or machine is said to be intelligent or have its own autonomy is as much a philosophical question as a scientific or legal one. We are reminded of this in a wide variety of popular culture references – think Blade Runner or the Matrix. The connotation carried by the term ‘Artificial Intelligence’ is that an AI program is capable of exercising autonomous human thought or emotion.

However, the reality is that AI in our current context is nothing more than a piece of code or programming written by a human author to perform certain processes. AI is used as a tool to automate and improve efficiency of processing and data analysis. It is important to note that most AI processes and algorithms require some level of human intervention or training to allow the program to operate as intended.

If this technology becomes more advanced, and as less human interaction is required for AI technology to perform their prescribed function, it may raise questions as to whether the current position at law is sufficient to deal with the issues which arise with developing technology.

Current legal position in Australia on IP Ownership by Artificial Intelligence

The Copyright Act 1968 (Cth) (Copyright Act) provides that the author of a literary, dramatic, musical or artistic work is the owner of any copyright which subsists in that work. (See section 35).

The Copyright Act provides that copyright subsists in an original work where the author is a ‘qualified person’, being a person who is an Australia citizen or a resident in Australia. By this reasoning, AI cannot be the owner of the works as it is neither a corporation nor a "person" as defined by the Act.

In order for a work to be sufficiently original to attract protection under the Copyright Act, there needs to be an exertion of human skill, creativity and ‘sweat of the brow’.[2] There is also a requirement for the creation of the work to involve some independent intellectual effort. [3]

In IceTV Pty Ltd v Nine Network Australia Pty Ltd [2009] HCA 14, the High Court considered that factual data contained in a compilation is not sufficient to attract protection under the Copyright Act. Copyright law is concerned with the originality of the way ideas or facts are expressed in a literary work, as opposed to the ideas or the facts themselves.

Historically, at common law, outputs of data which are generated solely by computer programs without any work or skill from a human creator have failed to meet the requirements for protection under copyright as they are not original literary works.[4] This is apparent in Telstra Corporation Limited v Phone Directories Company Pty Ltd [2010] FCAFC 149 where the lack of human intervention in the preparation of telephone directories by a database which outputs factual data through the use of a software program did not constitute an original literary work. The principles from this case apply to any output of factual data which is produced by a database without human work and skill.

Considering the above, it is clear that the current position in Australia is that copyright must have a human author. However, compare this approach with the law in respect of patents in Australia under the Patents Act 1990 (Cth).

The recent case of Thaler v Commissioner of Patents [2021] FCA 879 confirmed that it is possible for a patent to have an AI program listed as the inventor on a patent registration, notwithstanding that actual ownership cannot be vested in the AI. In that case, the Court considered that the owner and controller of the AI program would own any inventions made by the program and was therefore entitled to apply for registration of a patent which was invented by the AI program. While the AI can be listed as the inventor on the registration, it is not ultimately recognised as the owner given that it is not a legal entity.

It is important to note that the case of Thaler v Commissioner of Patents is currently subject to appeal and the view taken by the Federal Court may be subject to further consideration by the Courts.

Just because the law in Australia does not recognise ownership of intellectual property by AI, doesn’t mean that there are not potential issues which arise from developments in AI technology, both ethically and legally.

As AI improves and becomes more advanced in replicating human thinking, the law may need to reassess how it deals with ownership of IP to identify where IP should correctly sit or to clarify in statute whether IP created by AI is capable of constituting original works despite the law discussed above.

Our existing intellectual property laws were not written with AI in mind and as AI becomes more autonomous and requires less human intervention and training, it may be necessary for our IP laws to be revisited to account for the advances in technology.

Examples of AI and Intellectual Property

In order to appreciate some of the complexities and ambiguities which arise in this area, it is helpful to consider the different ways AI has been used and its potential use in a wide variety of industries.

It is reported that music producers are exploring the use of AI machines for music composition and production. Certain software programs can create unique drum patterns or melody patterns by analysing scores of music and recreating original samples. Other technology has been used to create original song lyrics by reviewing literary source material and creating new combinations of words which can be used as song lyrics.

Take for example, the 2016 song ‘Daddy’s Car’ which was produced by Sony Music researchers. The song was composed and written by an AI machine which had been trained to analyse and mimic the musical stylings of the Beatles. The final recording of the song required some editing and vocals by musician Bennoit Carre, and was therefore not completely free from human skill and ‘sweat of the brow’. It is clear that there is still sufficient creative ingenuity implemented by the musician themselves to finalise the musical work and the AI is merely a tool to assist them to produce the end result.

However, this technology is in its infancy and its reasonable to assume that as the technology develops there could be less and less human input required. This raises questions of ownership and originality when we are no longer talking about factual data generated by machines, but creative literary works developed by a program.

If, for example, a composition created solely by AI was taken to infringe upon another musician’s copyright, who should be responsible for that infringement?

Another example is the painting ‘The Next Rembrandt’ - a computer generated, 3D printed painting produced by AI using facial recognition from 346 paintings to create an artwork resembling the style of Rembrandt Hamenszoon van Rijn.

Other examples include AI programs writing dialogue to resemble the works of Shakespeare or Jane Austen. While these are done by human creators writing programs to analyse data samples and produce outputs, it is interesting to consider whether there is the possibility for programs to be trained to replicate creative writing processes of their own volition.

Legal developments overseas  

In jurisdictions such as the UK and New Zealand, the relevant legislation includes provisions for copyright ownership of works created by AI. For example, the Copyright, Designs and Patents Act 1988 (UK) (CDPA) section 9(3) provides that, in the case of a work which is computer-generated, the author shall be the person ‘by whom the arrangements necessary for the creation of the work are undertaken’. If the AI has been trained or instructed by a person to carry out a certain task, it will be that person who will be taken to be the author of the work.

Some U.S public interest groups, such as the Creative Commons Organisation, argue that works created by AI should remain in the public domain given the uncertainty of authorship, the differing levels of human involvement and the constant development of technology in this area.    

In the EU, the European Parliament has made recommendations to the effect that it should further consider the implications and legal impact of creating specific legal status for robots/AI and whether it is appropriate to apply electronic personality to robots and AI which make autonomous decisions and interact with third parties independently.[5] This in itself is a separate legal and ethical debate.

Ultimately no significant movement has yet been made by the EU in this sphere. However, it does demonstrate that governments all around the world are considering their policies and the legal and ethical implications of AI developments and what it will mean for a wide variety of law, including IP ownership.

Where to now for Australian IP Law?

The position in Australia is settled for now, and the requirement for human authorship remains paramount. What remains unclear is how jurisdictions around the world will grapple with further technological advancements and the policy considerations at play as the current laws become inadequate to deal with more complex AI which could operate independent from human interaction and exercise ‘human ingenuity’.

The World Intellectual Property Organization has commented that the approach adopted by the UK under the CDPA may be the most sensible approach to clarifying ownership of works created by AI, however it is unclear whether this is sufficient to address situations where there is no person who has made arrangements for creation of a work by the AI.[6] In that case, how far should the law go to trace the original ‘author’? Given that the current use of AI reflects use as a tool to achieve a specified outcome within specified parameters, it seems that the UK approach would be the appropriate approach to be adopted in the Australian context.

Authors: Jason Sprague & Jonathan Harris


[1] John McCarthy, “What is Artificial Intelligence?”, Stanford University, published 24 November 2004

[2] IceTV Pty Ltd v Nine Network Australia Pty Ltd [2009] HCA 14.

[3] Sands & McDougall Pty Ltd v Robinson (1917) 23 CLR 49 at 52 per Isaacs J

[4] Telstra Corporation Limited v Phone Directories Company Pty Ltd [2010] FCAFC 149

[5] European Parliament Resolution of 16 February 2017 with recommendations to the Commission on Civil Law Rules on Robotics, at Paragraph 59(f)