Artificial intelligence and copyright: ownership issues in the digital age - Copyright

In September 2015, the foundations of copyright law were shaken when animal rights group PETA sued British photographer David Slater on behalf of a monkey named Naruto to assert copyright over a 'selfie' taken by the monkey on the photographer's camera. ¹

Although the dispute eventually settled out of court, it raised unique issues about the nature of copyright ownership and challenged the foundational principle that creative works must be authored by a human to attract copyright. As we plunge deeper into the digital age, the concept of authorship is being further muddied in respect of non-human authors – that is, works created substantially or wholly through artificial intelligence (AI).

The COVID-19 pandemic has emphasised the need for societies, workforces and even entire nations to mobilise quickly to work, learn remotely and wholly embrace technology. In equal measure, it has emphasised society's commendable ability to do this, as we continue to bear witness to the many creative and collaborative outputs yielded through a time of mass isolation. Creative industries have been hit exceptionally hard by the pandemic, particularly in respect of live performances and television and film production.

In some fortunate cases, these industries have been able to adapt and shift from requiring a physical presence to inhabiting and collaborating in online spaces. As human-to-human collaboration is disrupted by the pandemic, we are seeing a rise in human-to-computer collaboration. As we look towards a post-pandemic world, we anticipate that the role of technology will be elevated even more so. With these drastic technological shifts in mind, this article explores the impacts of AI on copyright law, particularly in relation to traditional authorship constructs.

The Australian context

In Australia, copyright subsists in original works which are authored by a qualified person, meaning an Australian citizen or a person resident in Australia. ² Similar requirements apply in countries all over the world. These global constraints are problematic in circumstances where works are not created through human intelligence but rather through 'artificial neural networks' or 'brain-inspired systems that are designed to imitate the way the human mind learns'. ³

Indeed, this issue has already been tested to some extent by Australian courts. In Telstra Corporation Limited v Phone Directories Company Pty Ltd, ⁴ the Federal Court found that telephone directories did not attract copyright because they were 'not the result of human authorship but ^were computer generated' and did not involve 'independent intellectual effort' by the human contributors. ⁵ Likewise, a decade on from this decision, artificial intelligence has come a long way, and computer-generated works are making vast improvements in terms of sophistication and even creativity. Through machine learning, computers are able to absorb vast amounts of creative data, analyse the features and patterns of this data and generate something entirely new.

AI in the arts

As a society, we have become increasingly comfortable with the idea that AI can be applied to produce non-creative outputs, such as data processing and analytics. We have even largely accepted the fact that, in some instances, AI can modulate or even replace the work of humans in particular industries. There has traditionally been resistance, however, to the notion that AI could parallel, or even come close to mimicking, the human imagination.

However, there are a number of examples that demonstrate the creative capabilities of AI.

For example, in 2018, 1 the Road by Ross Goodwin became 'the first novel written by a machine'. ⁶ On the journey from New York to New Orleans, the AI machine consumed data generated from Goodwin's Cadillac car (which he had outfitted with a surveillance camera, GPS, microphone and clock) and produced a manuscript in real time. While 1 the Road is a strong example of the creative capabilities of AI, it was achieved with some very human elements of production, direction and independent intellectual effort. As noted by American critic Connor Goodwin, "^Though Goodwin has surrendered creative license to the writing machine, he nevertheless created the machine and the rules by which it operates". ⁷

Paving the way for works like Goodwin's was a Japanese novel co-authored by an AI system in 2016, aptly titled The Day A Computer Writes A Novel, which passed the first round of screening in the Hoshi Shinichi Literary Award. The novel was generated by reassembling words and phrases deconstructed from a template novel authored by the human team working with the AI program. Although the team created a choice-matrix for the AI program to follow, the human input in the wording of the novel is arguably more evident than Goodwin's in 1 the Road, as demonstrated by the novel's closing paragraph (translated by a report in The Japan News):

"I writhed with joy, which I experienced for the first time, and kept writing with excitement. The day a computer wrote a novel. The computer, placing priority on the pursuit of its own joy, stopped working for humans."

The novel was entered in the prize two years after Hoshi Shinichi's daughter, Marina Hoshi Whytemade, allowed the inclusion of non-human authored entries. The fact that the judges were not told which works were authored by humans and which were not is an indication of the relative quality of the novel and the increasing maturity and flexibility of AI programs.

There are also a number of musical and artistic works which have been artificially created by non-human intelligence. For example, an AI machine developed by Facebook AI research, Rutgers University and the College of Charleston was able to generate artistic works after being trained on the characteristics of over 80,000 paintings from the 15^th to the 21^st centuries. ⁸ The AI software generated abstract art which rivalled human authors so much that some critics could not tell the difference. ⁹

In the music world, American singer-songwriter Taryn Southern used machine learning in 2018 to create I AM AI, the first album to be entirely composed and produced using AI. Then there's 'Alice', AI developed by Australian start-up Popgun in 2017. Alice was initially skilled in listening to notes played by a human on piano and responding with an algorithm-based progression, however, 'she' went on to compose her own piano, bass and drums together as a backing track for human vocals. ¹⁰ Over the past year, Popgun have also been teaching an AI to sing. ¹¹

These examples are by no means exhaustive and as AI continues to advance in leaps and bounds the list will only grow. The 'computer-generated' works contemplated by the Australian Full Federal Court in Telstra Corporation Limited v Phone Directories Company Pty Ltd pale in comparison to today's technological applications. Indeed, to apply decade-old principles to the copyright issues arising from these developments is problematic to say the least.

Commercial implications: testing the IP 'rewards' system

The copyright system has been designed to protect and incentivise human intellectual effort. It does this by creating a framework in which, among other incentives, authors are rewarded financially for their work when it is used by others, and those who seek to circumvent this system are penalised. This framework is destabilised by the concept of AI authors – if AI-generated works are not protected by copyright because they have not been created by a human author then, theoretically, it follows that they could be freely exploited by anyone. This could have a chilling effect on investment in AI systems to produce creative outcomes.

One way for lawmakers to deal with this issue is to attribute authorship to the creator of the AI system. Indeed, other Commonwealth jurisdictions such as New Zealand and the United Kingdom have updated their copyright legislation to reflect the increasing role of technology and AI in creative works. In both these jurisdictions, the author of a literary, dramatic, musical or artistic work that is computer-generated is deemed to be 'the person by whom the arrangements necessary for the creation of the work are undertaken'. ¹² This language is similar to the protections that exist under Australian law for creators of cinematograph films, being 'the person by whom the arrangements necessary for the making of the film were undertaken'. ¹³ However, it remains to be seen whether Australia will extend this concept of authorship to literary, dramatic and artistic works.

Even if Australia does move in this direction, there is a still a question of who the law would consider to be the person making the necessary arrangements for the work to be created – the maker or user of the AI program?

Some commentators have drawn attention to the complexities brought about by AI over other technological advancements. For example, "Microsoft developed the Word computer program but clearly does not own every piece of work produced using that software". ¹⁴ Rather, the copyright in those works is attributed to the user. However, the user's contribution in works generated through artificial intelligence is likely to be much less significant. For this reason, it would seem to make the most sense to attribute copyright ownership to the creator of the AI program. This would likely also have the flow on effect of stimulating the invention of, and investment in, creative AI systems.

In other jurisdictions, AI machines themselves have been given the status of creator. For example, music composing AI, Aiva (Artificial Intelligence Virtual Artist), recently became the first AI composer to be officially recognised by SACEM, the France and Luxembourg authors' rights society. ¹⁵ As a result, Aiva can now release music and earn royalties under its own name. This recognition signals a shift in attitudes towards the role of AI in the arts and helps to overcome problematic scenarios where computer-generated works are not protected against unauthorised use or reproduction simply because they do not have a human author. On the other hand, there may be some resistance to this idea, as the computer-generated author would not have expended the same emotional or intellectual effort / 'sweat of the brow' to produce the work, so why should users have to pay for it? The increasing role of AI in creative works creates further tension in the creator/user dichotomy.

Attributing authorship to the creators of AI systems may, however, be problematic from an infringement perspective. For example, if an AI system creates a work that is substantially similar to a pre-existing copyright work, will the creator of the AI system be responsible for that infringement? Is this an unintended consequence that goes to show just how much work legal systems have to do catch up to these technological advancements? This will surely be an area for further consideration in the near future.

What about moral rights?

A further consideration is how the law will deal with moral rights. Moral rights are distinct from economic rights under copyright law in that they attach to the individual author and cannot be assigned to another person, not even the copyright owner. These rights are an important element in the philosophical matrix of copyright law as they provide a sound legal acknowledgement of the integrity of the creative endeavour and help to further the objective to incentivise the creation of new works. ¹⁶ In this case, moral rights should theoretically attach to the individual human authors responsible for the creation of the AI system itself.

However, it will be interesting to see how this will play out in practice, particularly if and when these AI systems are used by people other than the creator. Will these users be required to attribute the creator each time they produce work via the AI system? What about instances where unsavoury creative works are produced using the AI system – will the creator's right against derogatory treatment be triggered? These are just some of the questions that come to mind.

Looking to the future: is Australia already behind?

Anxieties about the human authorship test in the Copyright Act are not new. In fact, over twenty years ago, the Australian Copyright Law Review Committee voiced concerns about the extent to which the Copyright Act in force at the time accommodated 'the increasing, indeed almost ubiquitous, use of computers in the creation of copyright subject matter'. ¹⁷

As outlined above, there are already a number of economic and ideological considerations for Australian lawmakers to consider, and this list will only grow as AI technology advances.

Other common law jurisdictions around the world have already taken action to adapt to these advances, and Australian lawmakers will need to move quickly to ensure that Australian copyright law can cope with the many complexities brought about by the role of AI in content creation, particularly as we move into a post-pandemic world.

Footnotes

¹Jason Slotkin, ''Monkey Selfie' Lawsuit Ends With Settlement Between PETA, Photographer', NPR, 12 September 2017, available here
² Copyright Act 1968 (Cth), s 32.
³ Luke Dormehl, 'What is an artificial neural network? Here's everything you need to know', Digital Trends, 6 January 2019, available here
⁴Telstra Corporation Limited v Phone Directories Company Australia Pty Ltd (2010) 264 ALR 617
⁵ Ibid, 621.
⁶ Connor Goodwin, 'A.I. Storytelling: On Ross Goodwin's 1 the Road', Bomb Magazine, 14 December 2018, available here.
⁷ Ibid.
⁸ Elgammal et al (2017) 'CAN: Creative Adversarial Networks Generating "Art" by Learning About Styles and Deviating from Style Norms', paper presented at the International Conference on Computational Creativity (ICCC), Atlanta (20-22 June 2017), available here.
⁹ Elgammal et al (2017) 'CAN: Creative Adversarial Networks Generating "Art" by Learning About Styles and Deviating from Style Norms', paper presented at the International Conference on Computational Creativity (ICCC), Atlanta (20-22 June 2017), available here.
¹⁰ Jacca-RouteNote, 'Is the future of pop music in artificial intelligence?', RouteNote, 6 November 2018, available here.
¹¹ Popgun Labs, 'Popgun – Vocals', YouTube, 21 July 2019, available here.
¹² Copyright Act 1994 s 5(2); Copyright, Designs and Patents Act 1988 (UK) s 9(3).
¹³ Copyright Act 1968 (Cth) s 22(4).
¹⁴ Andres Guadamuz, 'Artificial intelligence and copyright', WIPO Magazine, October 2017, available here.
¹⁵ AI Business, 'Aiva is the first AI to Officially be Recognised as a Composer', 10 March 2017, available here.
¹⁶ Australian Law Reform Commission (2013) Copyright and the Digital Economy, Report No. 122, 7.
¹⁷ CLRC, Parliament of Australia (1999) Simplification of the Copyright Act 1968 — Part 2: Categorisation of Subject Matter and Exclusive Rights, and Other Issues, 47–8 [5.10].

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.