IP legislation often finds itself struggling to plug gaps in the law caused by the rapid pace of technological change, and the state of the law surrounding ownership of AI-generated products is no different. In the first article of this series, we considered how current Australian patent and copyright law frameworks would deal with questions of AI ownership for AI-generated IP.
The question that arises for the next generation of lawmakers, and the subject of this article, is then: should AI machines be recognised as owners of IP under the law? We consider both the advantages and disadvantages of such protection below.
Arguments for AI ownership
(1) It incentivises growth in the AI industry
Recognition of at least some degree of AI ownership may be a way to afford IP protection to AI-generated outputs, which simply do not have identifiable human authors or inventors. Allowing such outputs to fall in the public domain may disincentivise investment in the area and slow down the growth of such industries.
(2) It correctly identifies the true inventor / author of the IP
Some argue that AI machines ought to be recognised as the true inventor and author, as it is the machines, rather than the human programmers, with the expertise to contribute to the ‘inventive step’ and ‘independent intellectual effort’ required for inventorship in patents and authorship in copyright respectively. For example, Dr Stephen Thaler, the creator of DABUS AI, points out that he would not have had the expertise to create food containers based on fractal geometry without the work of DABUS AI.
(3) New types of IP protection better accommodate modern concepts of IP
AI-generated products do not fit into current conceptions of IP law, so the best way forward may be to create an entirely new type of protection, like the EU has done in respect of database rights.
Arguments against AI ownership
(1) IP protection is not necessary for AI creations
Historically, the purpose of IP protection has been to promote sciences and arts by incentivising innovation and creativity through patent and copyright protection. However, as machines do not require such incentives to continue generating products, there is arguably no reason to grant ownership rights to the AI machines themselves.
(2) It may be detrimental for the actual humans behind the AI work
Recognising AI machines as IP owners may result in the denial of IP rights or recognition for the human contributions still required to reach the final AI-generated output.
For example, AI artist communities see AI merely as a tool to create their artwork, just as a traditional artist would with ink and paint. To attribute ownership to a machine ignores the creative direction, design and innovation put in by the artists.
(3) AI machines don’t have the requisite legal personality to enforce IP rights
IP legal systems provide owners with a suite of rights to be enforced, protected and exercised. In order for such systems to work however, IP owners need to be able to exercise such rights, for example, entering into contracting agreements, commercialising products and taking legal action. AI systems do not have the requisite legal personality to do so, and so wouldn’t be able to exercise such rights, even if granted them.
(4) Potential conflict between owners and other stakeholders
A product is rarely made by one individual, but rather is developed by a team, each contributing different skills and expertise. Granting ownership only to an AI machine ignores the complexities behind the network of human contributions leading up to the end product.
For example, the Obvious AI team admits that it did not invent the code for the painting generated by Obvious AI, but rather obtained it from AI artist Robbie Barrat via discussions from an open source website forum. This scenario emphasises the potential complexities around allocating ownership. Should ownership belong to the code programmer (Robbie Barrat), the product creator (the Obvious AI machine), the product manufacturer (the Obvious AI team), or all three?
Conclusion and takeaways
While this area of law will likely be greatly challenged with the rapid growth of the AI industry, AI ownership of IP is still currently not recognised under traditional mechanisms of IP protection in many countries, including Australia. For companies working with products generated via automation processes, keep in mind that you may face inventorship and authorship issues under the current legal frameworks.
For the time being at least, robust and well-structured contracting arrangements and licences between the different stakeholders in the creation process are the most important means of protecting your IP rights.
The authors wish to thank summer clerk Ying Yi Lim for her contribution to the research and drafting of this article.
 D Lim, ‘AI & IP Innovation & Creativity in an Age of Accelerated Change’ (2019) Akron Law Review 52(3) 836-847.
 Some commentators argue that the best legal personal to allocate ownership rights to in order to maximise economic efficiency, is that of human product manufacturers. This is as such manufacturers are able to best commercialise such products, as compared to the AI program coders and the AI machine itself.