Introduction

In recent years, artificial intelligence or robots have been creating contents on their own. The famous Edmond de Belamy, which was sold for an unexpectedly high price of US$432,500 at auction, was an artwork created using artificial intelligence (“AI”). Shimon, the singer-songwriter robot empowered by artificial intelligence which sings and writes its own lyrics and composes its own melodies, is about to go on tour to promote its first album. With an increasing amount of funds being invested in such technologies and the potential commercial benefits behind, crucial issues arise and have to be dealt with – can AI-generated contents be considered as copyrighted works? And if so – who should be the owner of the copyright?

Can AI-generated contents be
considered as copyrighted works

In Hong Kong, copyright can subsist in literary works, musical works, dramatic works, artistic works, etc. As Hong Kong upholds the “Sweat of the Brow” principle in determining originality of works, originality is not concerned with whether the work is inventive, novel or unique, but whether the work is originated from the author and is not copied from elsewhere. The author must have exercised the requisite “labour, skill or effort” in producing the work. This can be contrasted with the copyright law in other jurisdictions such as the European Union which requires the work to be the author’s own intellectual creation rather than merely a work of “labour, skill or effort”. In the United States, the requirement is that the work should be considered original plus at least some minimal degree of creativity. The PRC similarly requires originality for a creation to be considered as a “work” according to Article 2 of the Regulations for the Implementation of the Copyright Law.

Originality of AI-generated works

No doubt, the computer program designed for the AI robot by the author is protectable under the copyright law. However, how about AI-generated works? The answer lies on the issue as to whether the works are “original”. When considering this element, one must not separate the work into dissected parts and consider whether or not there is copyright in each of the parts (Ladbroke (Football) Ltd v William Hill (Football) Ltd [1964] 1 WLR 273 (HL)). The work must be assessed as a whole. It was held in the Hong Kong case of Tai Shing Diary Ltd v Maersk Hong Kong Ltd [2007] 2 HKC 23 that “an author might draw on existing material, so long as more than negligible or trivial effort or relevant skill and judgment had been expended in the creation of the work. The standard required was a low one, but the effort must not be so trivial as to be characterised as a purely mechanical exercise. The skill and effort protected was not only that expended on its manner of presentation, but of collecting, selecting, arranging and presenting the available information in an intelligible manner”. In general, AI generates works under prescribed programs and through the application of machine learning algorithms provided by programmers or designers. For example, Shimon the robot was provided with datasets of 50,000 lyrics from jazz, prog rock, and hip-hop, and was “trained” to imitate the workings of the human brain in processing data and creating patterns for use to create melodies and lyrics, i.e. using deep learning. While it is arguable that the process of generating AI-generated contents is “purely mechanical”, one cannot deny that “skill and judgment” are the two main elements forming part of the creation of AI-generated works. It therefore remains arguable as to whether AI-generated works satisfy the basic requirement of “originality”. In fact, as the application of AI continues to develop and is expected to become more advanced and complicated in the near future, it is inevitable that the issue of “originality” will be open to more debate.

Who should be the owner?

If AI-generated contents are to be considered as copyrighted works, a logical question arises as to who the copyright owner should be. Section 178 of the Copyright Ordinance (Cap. 528 of the Laws of Hong Kong) provides that a work qualifies for copyright protection if the author is a natural or legal person. Similarly in the PRC and many other jurisdictions, only natural and/or legal persons can enjoy copyrights.

It necessarily follows that under the existing framework of these jurisdictions, the robots, programs or computers that generate such contents cannot be the copyright owner. Even though Hong Kong upholds the “Sweat of the Brow” principle in determining originality of works such that robots, programs or computers may on the face of it be qualified as the copyright owner, they fail to fulfill the requirement under section 178 of the Copyright Ordinance as they are not natural or legal persons. Furthermore, affording copyright protection to robots, programs or computers may not achieve one of the ultimate objectives of copyright law as robots cannot be incentivised to create. As robots are not legal persons, practical considerations arise as they cannot sue or be sued. It is difficult to recognise AI as a legal person given its current stage of development.

To fill the vacuum of authorship so as to avoid the AI-generated contents being left in the public domain, alternative owner(s) must be identified. An option would be the programmer or the designer of the robot. Providing copyright protection over AI-generated works could understandably incentivise programmers to write the underlying algorithms. However, as AI evolves over time, it is probable that the original programmer would not be able to foresee its subsequent development and the works it generates. Another plausible option would be the user of the robot which creates the contents, as the user would be the one providing inputs to the AI. A similar problem in connection with this option is that it is difficult for the user to contemplate the end product using AI. What is more, AI programs can be used by anyone and generate the same contents using the same keywords or inputs. One has to consider what exactly copyright law seeks to protect – works created by human intelligence, or the intellectual activity of choosing keywords or inputs.

Notably, some jurisdictions such as the United Kingdom and New Zealand have already taken the lead to make changes to their copyright legislation, recognising authorship in AI-generated contents as belonging to the person “by whom the arrangements necessary for the creation of the work are undertaken”. It means that the person creating the program will bear the fruits by holding copyright to any works which the program may produce. This change is welcomed, as businesses are provided with greater certainty that their investments in AI programs will be protected.

The way forward

The above merely illustrated two of the many issues arising out of AI-generated works. It is clear that the existing legal framework in Hong Kong offers limited, if any, protection to such AI-generated works. Although it is imperative that changes have to be made to the copyright law to accommodate the development of AI and currently there is no official registry in Hong Kong for registration of copyright works, authors of the relevant works could endeavour to maximize the scope of protection for their works by means of a clear declaration that the work is protected under copyright law, keeping a record of supporting evidence of the evolution of ideas, marking footprints or watermarking and/or using a “©” symbol, etc. as much as possible.