Abstract The author Jani Ihalainen is a Master of Laws graduate who currently works at an international law firm in London. He has a particular interest in IP and its convergence with new-age technologies. This article This article aims to set out the current legal status of artificial intelligence and works created by it in terms of copyright in the UK and EU. The legislative arms of most countries have not specifically dealt with this future problem, particularly as these works would potentially not be covered by copyright, and if they are, would potentially create monopolies for those who would create works en masse. The legal context for works created by artificial intelligence is unclear, even if parallels are drawn between works created using computers there is no clear status for these works in most copyright regimes. Should copyright protection be assumed under current legislation, this could lead to unintended consequences and a new age of copyright trolling. The article’s opens up the conversation around copyright and artificial intelligence, highlighting possible issues and solutions through other jurisdictions outside of the EU and common law countries. Potential legislation could be considered through this conversation, which will alleviate the growing pains from the coming era of artificial intelligence. 1. Introduction Sci-fi authors have long offered us a glimpse of a future where machines are capable of original thought and action without human input. For years lack of sufficient computer processing power stopped this becoming reality, until cloud computing came along and suddenly tipped the balance. It now feels like we are in a corporate space race to develop sophisticated artificial intelligence (AI) and machine learning systems, with ethicists, lawyers and regulators sprinting to understand how the developments will affect society. This article addresses one very specific impact of the growth of AI capable of creating original works: who will own the copyright in these works? 2. The technology What actually amounts to AI is still being debated, and no established definition has been set, but it has been described as synthesized intelligence1 through both computer software and hardware, essentially ‘… construct[ing] computer programs … capable of exhibiting intelligence’.2 Definitions of intelligence vary widely between academics and researchers, but could be argued as being the ability to ‘… reason, achieve goals, understand and generate language, perceive and respond to sensory inputs, prove mathematical theorems, play challenging games, synthesize and summarize information, create art and music, and even write histories’.3 While creative AI is still in its formative years, there are currently examples of its application in the arts. Flow Machines is an AI system that creates music either independently with no human input or in collaboration with human authors.4 It breaks down music styles into computational objects and strings them together to make songs in that style. The technology company DeepMind has also developed WaveNet, a deep generative model of raw waveforms, which can mimic human speech in a very natural-sounding way.5 Arguably this could lead to AI’s ability to sing, which, paired with Flow Machines, could produce music with vocals independent of any human input at all, and with a great deal of complexity, potentially even mimicking famous, well-established musicians. The technology is by no means restricted to only creating music. Future University Hakodate’s AI program wrote a novel called The Day a Computer Writes a Novel, which passed the first round of judging for the Hoshi Shinichi Award.6 The AI program did not create the works without any human involvement, as the researchers selected words, sentences and set parameters for the novel, but did create the ultimate work submitted for the award. AI has also challenged the art world, where researchers from both Rutgers University and Facebook’s AI lab have created an AI that can make pieces of art using a generative adversarial network, where neural networks create and judge each other’s work, looping until a desired work has been finished.7 The neural networks are trained through viewing thousands of paintings, even being capable of determining what humans would class as a great piece of art or not, and honed to create art that fits a particular style, such as Cubism or Rococo. These types of AI could create on-demand content based on users’ selections, allowing them to pick themes and styles for that content based on their desires at any given time. The AI would then create new works for them, and with no time limitations like with human authorship, the potential number of works that can be created could be near infinite. 3. Computer v human—who is the ultimate author? The UK Copyright, Designs and Patents Act 1988 (CDPA) added a provision that allowed for the authorship of computer-generated works (where the work is generated by a computer in such that there is no human author)8 to vest in the ‘the person by whom the arrangements necessary for the creation of the work are undertaken’.9 This provision could be problematic in the light of possible future developments in AI, but will be discussed more in depth later in this article. Current case law has discussed human authorship either through the use of computers or using software to generate works, much like in the CDPA. The case of Express Newspapers Plc v Liverpool Daily Post & Echo Plc10 concerned the copying of a grid used in a newspaper lottery, which was ultimately created using computer software to generate the grids. In the case the High Court of England and Wales saw that the computer was no more than a tool through which the author created the works,11 akin to an artist’s pen or a typewriter. Arguably, this is correct, as the computer and the programs employed act as a conduit to manifest the expression of the human author, rather than create the works themselves without the human’s input. In a more recent case, Nova Productions Ltd v Mazooma Games Ltd,12 the High Court had to determine whether computer game elements that were generated using bitmap files made by the programmer setting out pieces of the elements in the game were indeed the authorship of the programmer. The court decided that the composite frames generated by the computer program were the authorship of the programmer as ‘… he devised the appearance of the various elements of the game and the rules and logic by which each frame is generated and he wrote the relevant computer program’.13 He therefore ‘… is the person by whom the arrangements necessary for the creation of the works were undertaken and therefore is deemed to be the author’,14 following the current provisions of the CDPA. As discussed above, this interpretation can be problematic, especially in the light of AI, as merely the making of the ‘arrangement necessary’ for the creation of the works could lead to a virtual monopoly in AI-created works. A programmer or a company that designs an AI (including all of the other roles involved in the development of the software and infrastructure) that could create, for example, musical works according to a few criteria set by the end user, would arguably be making the ‘arrangements necessary’ for the creation of those works, and could potentially own the rights in a near-infinite amount of copyright protected musical works. Any future legislation around AI and copyright would therefore have to distinguish between computer-created works and works created with the assistance of computers, seeking to protect the interests of active, computer-assisted creators, but limit the rights in passive AI-created works in the absence of clear and substantial human input in those works. 4. A foreign perspective The US Copyright Act of 1976 requires for a work to be protected by copyright for it to have been created by an ‘author’.15 Unhelpfully the legislation does not define who an ‘author’ is, which many other common law countries do, but recent cases have looked at human and non-human authorship. In Naruto v Slater,16 the US District Court of the Northern District of California faced the question of animal ownership in photographic works, where a monkey had used a photographer’s camera to take a picture of itself. The decision refused the monkey’s claim for authorship, as the copyright legislation largely speaks of a ‘person’ involved in the creation of the work, and that for a work to qualify as a copyright protected work it has to be ‘… created by a human being’.17 Although PETA, representing Naruto, appealed the decision, the case has since settled out of court.18 The case follows the US Copyright Office’s Compendium,19 which expressly excludes non-human authorship. This would undoubtedly extend to works created by AI, although it is arguable that if there is significant human input in the AI’s work the human could have authorship. It is likely that there would be no protection for works that would be created by AI in the US, while the use of computers by a human to create the same would most likely be protected. The US Copyright Office has highlighted this as a future issue as far back as 1965;20 however, it seems to have been ignored in the following 50 years. The Canadian Copyright Act sets the requirement that the author of a work must be either a citizen or subject, or a person ordinarily resident in a treaty country.21 There is no provision that deals with computer-generated works, nor any cases dealing with the same; however, one can argue that copyright would only apply to works created with the assistance of computers (ie as tools as discussed above), and would not apply to computer-generated AI creations. There is no equivalent provision in Canadian copyright legislation for making the ‘arrangement necessary’ for the creation of computer-generated works; however, there is one case that could extend copyright to AI created works. In Apple Computer Inc v Mackintosh Computers Ltd22 the Supreme Court of Canada afforded copyright protection to machine code that was embedded in a computer’s Read-Only Memory (ROM). This could extend copyright protection not only to the codes programmed in complex languages or on advanced artificial intelligence platforms, but also to the resulting object code.23 In Australia the Copyright Act 1968 speaks, similarly to the UK, of a ‘qualified person’24 as a requirement as the author of a work. This clearly excludes non-human authorship, and would leave AI-created works beyond the remit of copyright, which was confirmed by the Federal Court of Australia in Telstra Corporation Limited v Phone Directories Company Pty Ltd,25 where the automatic computer generation of phone directories was deemed to not be the authorship of anyone.26 In a more recent case, Acohs Pty Ltd v Ucorp Pty Ltd,27 the Full Federal Court of Australia considered that the underlying HTML code for information sheets generated by a computer program would have no author, and therefore no copyright protection. This excluded both the creators of the program and the individuals inputting the information into the program to create the sheets. Australian legislation does not have provisions for making the ‘arrangement necessary’ for the creation of the work, and doubts the existence of copyright in works created by AI, even with human input. It has been suggested that this could be implemented in Australia to combat this issue,28 and AI was not addressed in the recent Australian Productivity Commission’s report29 or the Government’s response on reforms.30 Under EU legislation AI authorship seems equally doubtful. In the case of Infopaq International A/S v Danske Dagbaldes Forening31 the Court of Justice of the EU (CJEU) determined that copyright only applies to works that reflect ‘… the author’s own intellectual creation’.32 Following Infopaq, the EU courts have cemented the concept of ‘the author’s own intellectual creation’ to mean works where ‘… its author expresses his creative ability in an original manner by making free and creative choices … and thus stamps his “personal touch” [on the works]’.33 Clearly AI would fail the above test, as it would not be classed as an author, and the works it creates would not be considered original creative works and afforded copyright protection. With the above in mind, there seems to be a very slow push in the EU for developing the law around copyright to suit a more automated, AI-operated future. The EU Parliament’s Legal Affairs Committee’s study on European civil law rules in robotics34 has considered this problem, and recommended the adjusting of copyright in the light of autonomous robots’ new/future abilities,35 which can be argued to apply equally to AI even in the absence of a physical robotic entity. Their recommendations in the draft report36 demands criteria for ‘own intellectual creation’ for protectable works produced by computers or robots.37 What clearly stands out are the two issues that future legislation will have to tackle in this area: copyright ownership and enforcement. The courts could also prove an avenue for addressing the issue in the immediate future, as they have for example in Nova Productions; however, tailor-made and targeted legislation would provide certainty and a specific remedy for a very important issue going forward. As AI creation entails a great deal of monetary input, effort and independent creation as a part of a wider team of developers, who the ultimate owner of the rights is needs to be firmly established. Once the ultimate owner has been decided, enforcement of any rights in the works would have to be restricted, duly rewarding developers’ efforts, but allowing for others to create works (with or without human intervention) and to deter a very wide-reaching set of rights as currently afforded by copyright. Another consideration is whether AI should have rights itself, or where the rights should lie if the material produced has absolutely no human intervention (or planning) attached to it. In its draft report the EU Commission has suggested that we ‘creat[e] a specific legal status for robots, so that at least the most sophisticated autonomous robots could be established as having the status of electronic persons with specific rights and obligations … and apply electronic personality to cases where robots make smart autonomous decisions or otherwise interact with third parties independently.’38 While the report discusses the matter of robotics, the principles could be just as well extended to AI. Electronic personhood could afford them some rights in the works created, but who would be the beneficial owner of those rights (eg the manufacturer or the purchaser of a robot) or the proceeds would still remain unclear. It has been suggested that the owner of the AI should be awarded the rights;39 however, how the AI could use or enjoy the proceeds is unclear and a possible further ethical issue. One of the few nations forging ahead in the debate for AI legislation in copyright is Japan, whose government has been considering a new protection regime for non-human-created intellectual property.40 According to reports41 this new legislation would seek to limit the protection of AI-created works, treating them akin to trade marks and realigning their protection to unfair competition rather than a broader expression-based scheme. The particular works’ degree of protection would also be tied to its popularity, similarly to the concept of goodwill, leaving less obscure or popular works with little or no protection, safeguarding from a flood of works with the sole aim of benefiting through copyright enforcement. The works would be owned by the individual or the company that had created the AI in question. 5. Who should be the author? The question of authorship, as can be seen from the above, is incredibly tricky, and there is no clear-cut answer. Arguably the law should shift from an author-centric approach to one that protects individual works in isolation, much like trade marks, as illustrated by the possible Japanese approach. Not only would this allow for more flexibility in terms of the level of protection, but also mitigate against the mass production of works simply for the enforcement of rights in unknown works for a quick licence payment. Should an author be required for AI-created works, one could follow the lead of the EU in the report on robotics that suggests the creation of electronic personhood or even a specific set of rights for AI to own the rights in the works that it creates. One option that has been raised is to treat the works made by AI as works for hire,42 affording the rights to the person who has commissioned the AI to create a particular work. While this would be the most straightforward option, it could still lead to issues where companies could commission works and saturate the market, potentially even wielding the masses of works as a sword against other producers, including competitors. Jani McCutcheon has also suggested43 that AI created works could also benefit from sui generis rights, similar to database rights, therefore not requiring an author at all. The recognition of the investment in the creation of the works through AI would allow for them to be protected, but could still cause issues as has been discussed above. 6. The future There is a vacuum left between copyright and the developing world of AI. Much like during the emergence of the internet, the law seems to be playing catch-up, which leads to potentially negative outcomes given the rapid development and implementation of AI that is happening at the moment. Inaction in this area can leave creative industries vulnerable to the creation of copyright protected works by sophisticated AI, but also leaves the protection of creations by the aforesaid AI in limbo, not offering enough incentive to develop these platforms in the short term. As said above, any possible legislation would have to balance the interests of technology developers, but also safeguard those who do not use passive, computer-generated creation only. Since there seems to be very little impetus to create these new legislative schemes, the future of the law in this area seems very murky, but the efforts of the Japanese and the EU undoubtedly could lead the way for the future. Footnotes 1 Peter Stone et al., ‘Artificial Intelligence and Life in 2030’, One Hundred Year Study on Artificial Intelligence: Report of the 2015–2016 Study Panel, Stanford University, Stanford, CA, September 2016. Available at: http://ai100.stanford.edu/2016-report (accessed 22 January 2018), 13. 2 Herbert A. Simon, ‘Artificial Intelligence: An Empirical Science’, Artificial Intelligence 77, no. 2 (1995), 96. 3 Stone et al., ‘Artificial Intelligence and Life in 2030’, 13. 4 See http://www.flow-machines.com (accessed 22 January 2018). 5 ‘WaveNet: A Generative Model for Raw Audio’, see: https://deepmind.com/blog/wavenet-generative-model-raw-audio (accessed 22 January 2018). 6 ‘An AI-Written Novella Almost Won a Literary Prize’, see: https://www.smithsonianmag.com/smart-news/ai-written-novella-almost-won-literary-prize-180958577/ (accessed 22 January 2018). 7 ‘Artificially Intelligent Painters Invent New Styles of Art’, see: https://www.newscientist.com/article/2139184-artificially-intelligent-painters-invent-new-styles-of-art/ (accessed 22 January 2018). 8 Copyright, Designs and Patents Act 1988, S. 178. 9 Ibid, s. 9(3). 10  1 W.L.R. 1089. 11 Ibid, 1093. 12  E.M.L.R. 14. 13 Ibid, 105. 14 Ibid. 15 17 U.S. Code § 102; with the rights vesting in the author of the work only, 17 U.S. Code § 201. 16 Naruto v Slater, case no. 15-cv-04324-WHO (N.D. Calif. 2016). 17 Ibid. 18 ‘Settlement Reached: “Monkey Selfie” Case Broke New Ground for Animal Rights’, see: https://www.peta.org/blog/settlement-reached-monkey-selfie-case-broke-new-ground-animal-rights/ (accessed on 12 December 2017). 19 U.S. COPYRIGHT OFFICE, COMPENDIUM OF U.S. COPYRIGHT OFFICE PRACTICES § 101 (3d ed. 2017). 20 ‘Sixty-eighth Annual Report of the Register of Copyrights for the Fiscal Year ending June 30, 1965’, see: https://www.copyright.gov/reports/annual/archive/ar-1965.pdf (accessed 21 December 2017). 21 Copyright Act, RSC 1985, c C-42, s 5. 22  2 RCS 209. 23 ‘Intellectual Property and Artificial Intelligence’, see: https://www.lexology.com/library/detail.aspx?g=12cde3b1-80f9-4034-bd93-090d416ec34c (accessed 23 January 2018). 24 Copyright Act 1968, s. 32. 25 Telstra Corporation Limited v Phone Directories Company Pty Ltd  FCA 44. 26 Ibid, 179. 27  FCAFC 16. 28 Jani McCutcheon, ‘The Vanishing Author in Computer-Generated Works: A Critical Analysis of Recent Australian Case Law’, Melbourne University Law Review, 36 (2013); UWA Faculty of Law Research Paper No. 27, see: https://ssrn.com/abstract=2297192 (accessed 22 January 2018). 29 ‘Productivity Commission Inquiry Report on Intellectual Property Arrangements’, see: http://www.pc.gov.au/inquiries/completed/intellectual-property/report/intellectual-property.pdf (accessed 22 January 2018). 30 ‘Australian Government Response to the Productivity Commission Inquiry into Intellectual Property Arrangements’, see: https://www.industry.gov.au/innovation/Intellectual-Property/Documents/Government-Response-to-PC-Inquiry-into-IP.pdf (accessed on 22 January 2018). 31 Infopaq International A/S v Danske Dagblades Forening, Case C-5/08, ECLI:EU:C:2009:465. 32 Ibid, para 37. 33 Football Dataco Ltd and Others v Yahoo! UK Ltd and Others, Case 6-604/10, ECLI:EU:C:2012:115, 38. 34 ‘European Civil Law Rules in Robotics’, see: http://www.europarl.europa.eu/RegData/etudes/STUD/2016/571379/IPOL_STU(2016)571379_EN.pdf (accessed 21 December 2017). 35 Ibid. 36 ‘Committee on Legal Affairs, Draft Report with recommendations to the Commission on Civil Law Rules on Robotics’, see: http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML%2BCOMPARL%2BPE-582.443%2B01%2BDOC%2BPDF%2BV0//EN (accessed 21 December 2017). 37 Ibid, 21. 38 Ibid, 23. 39 Filipe Maia Alexandre, ‘The Legal Status of Artificially Intelligent Robots: Personhood, Taxation and Control’, see: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2985466 (accessed on 22 January 2018), 24. 40 ‘Intellectual Property Strategic Program 2016’, see: http://www.kantei.go.jp/jp/singi/titeki2/kettei/chizaikeikaku20160509_e.pdf (accessed on 21 December 2017). 41 ‘Japan Eyes Rights Protection for AI Artwork’, see: http://asia.nikkei.com/Politics-Economy/Economy/Japan-eyes-rights-protection-for-AI-artwork (accessed on 21 December 2017). 42 Annemarie Bridy, ‘Coding Creativity: Copyright and the Artificially Intelligent Author’ Stanford Technology Law Review, 5 (18 July 2011), 1–28, 26. 43 McCutcheon, The Vanishing Author in Computer-Generated Works: A Critical Analysis of Recent Australian Case Law (n 28), 965. © The Author(s) 2018. Published by Oxford University Press. All rights reserved.
Journal of Intellectual Property Law & Practice – Oxford University Press
Published: Mar 6, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera