The rapid advancement of generative artificial intelligence (“Gen AI” or “AI”) and its integration into modern life is gaining ground. Gen AI can create realistic and high-quality text, images, video, audio, music and more. It is the mainstay of large language models, which can generate content within seconds and exhibit humanlike creativity. The growth of Gen AI raises important questions as to whether AI-generated content can be accommodated within Sri Lankan copyright law as set out in Part II of the Intellectual Property Act, No. 36 of 2003 (the “IP Act”). While the creation of literary, artistic, musical and scientific content with the aid of technology is not new, what is unprecedented is the ability of AI to produce creative content, which until very recently, was largely, if not substantially, the preserve of human creativity.
As explained below, Sri Lankan copyright law is centred around literary, dramatic, musical or artistic works originating and resulting from the appropriate skill, labour and judgment of a human author. In the case of such “authorial” works, many of the rules of copyright are organised around the concept of human authorship. AI-generated content in the form of text, images, audio, video and music, however, sit uncomfortably with the concept of human authorship. Indeed, such relatively authorless AI-generated output is – in copyright terms – an enigma in that it lacks an identifiable human author who has sufficiently contributed to its creation. For that reason, the IP Act may need to be revisited to designate an appropriate author or owner of computer generated content including AI-generated output (e.g. the developer or user of the AI system) perhaps in the same way that a producer is regarded as the first owner of audiovisual works and sound recordings, and broadcasting organisations are given exclusive rights in relation to broadcasts.
The Requirement of Originality
The requirement of “originality” is vital for a certain “work” to obtain copyright protection. The IP Act protects literary, artistic and scientific works, which are “original intellectual creations”, though of course such protection is conferred on works “irrespective” of their “content, quality and purpose”.
Though the words “original intellectual creations” suggest a more demanding standard entailing the expression of “free and creative choices,” the Sri Lankan courts have been content to follow the traditional English law tests namely, that: (1) the works originate from the author and must not be “slavish” copies of existing works; and (2) the resulting works must be the product of the author’s “independent efforts” and evince sufficient “skill, labour and judgment” on his or her part.
At the very least, the requirement of originality requires independent and “substantial” human effort in the creation of a work, even if the usefulness, inventiveness, aesthetic merits, quality and value of the work are irrelevant. On the one hand, it is doubtful whether an AI-generated output dictated by complex but mechanistic algorithms and devoid of any meaningful human input – save for mere user prompts, which are intended to obtain a desired result – can satisfy this requirement of originality. Copyright law will refuse to recognise originality where the human skill, labour or judgment in relation to any work is trivial or insignificant. On the other hand, AI-generated content involving substantial creative human input, which has the effect of individualising or shaping the final output of any AI-generated content, may yield the requisite skill, effort and judgment needed to satisfy the requirement of originality.
For clarity, the IP Act does protect specific “entrepreneurial” works without the need for originality. Such works include sound recordings and broadcasts, which are protected without any requirement of originality. They are protected regardless of whether the statutorily designated owner has exercised any skill, labour or judgment in their creation. It may well be that the IP Act should be revisited to protect AI-generated content in the same way even though the creative subject matter generated by AI is “content” rather than “signal” based.
Authorship of Works
As noted above, many of the concepts and rules of copyright law in relation to authorial works (i.e., literary, dramatic, musical and artistic works) are organised around the existence of an “author”.
Under the IP Act, the initial owner of any copyright subsisting in any literary, dramatic, musical, artistic or scientific work is generally the “author who created the work”. Specifically, an “author” means the “physical person who has created the work”. Therefore, any qualifying literary, dramatic, musical or artistic work must be created by a human being. Accordingly, in the case of AI-generated content, there is doubt as to the existence of any identifiable human author, especially if the resulting AI-generated output is produced with or uninfluenced by any substantial or material human input. There can equally be no question of AI being the “author” of any resulting work: the AI being a tool or technical aid such as a pen or camera.
Likewise, several other concepts and provisions of the IP Act are linked to the existence of a human author. For instance, the requirement of originality signifies the link between the author and the work. That requirement simply means that a human author has exercised his or her skill, labour or judgment in creating a work or in other words that the work is the author’s own intellectual creation.
Similarly, while the exclusive “economic rights” attaching to any work may be vested in another person (whether natural or legal), the “author” is given – waivable – “moral rights” that are personal to the author, and which are transmissible only upon his or her death. Those moral rights include: the rights to have the author’s name indicated prominently on any copies of any work; the right to use a pseudonym in the alternative, and the right to object to any distortion, mutilation or other modification of, or other derogatory action, which would be prejudicial to his or her honour or reputation. In addition, the duration of copyright is calculated by reference to the “lifetime” of the author.
Indeed, the applicability of Part II of the IP Act particularly with respect to the protection of literary, artistic or scientific works hinges on the nationality or territorial connection of the author or to the initial publication of his or her work either in Sri Lanka or the territory of any other State that is a signatory to the Berne Convention or any other international agreement on copyright to which Sri Lanka is a party.
Therefore, the existence of a human author is a fundamental organising principle in copyright law and leaves open the risk that in the case of AI-generated content, no copyright would subsist for want of any identifiable human author who creates the work.
In the English case of Express Newspapers v Liverpool Daily Post & Echo Plc, computer-generated bingo sheets – comprising letter sequences in grids – were held to qualify as literary works, and their author was the programmer of the computer. This case was, however, decided before the express protection of computer-generated works under British law. The court held that the sequences entailed a human programmer expending sufficient skill and labour – in programming the computer, running and checking the results – and in this respect, the computer was a tool that carried out the programmer’s instructions in the same way that a pen was an instrument to produce written works. That case, however, was decided in 1985 at a time when the computer programmer wielded significant control in the input and final output of traditional computer systems. Gen AI, in comparison, can create or synthesise sophisticated content with minimal human involvement or input.
Other difficult and less likely possibilities include either the developer or user of the AI system being regarded as the “author” or “owner” of any work, or the developer and user qualifying as joint authors assuming that the requirement of collaboration can be met. Whether this is so will be highly fact specific and depend on the extent to which the relevant human intervention materially influences the way in which the resulting AI-generated output is expressed. This is because copyright law only protects the originality of the way in which a work is individualised and expressed (e.g.: the way in which an artist has chosen to paint a sunset as distinct from the idea of a sunset itself and a photographer’s choice over the angle, focus, lighting and position), as opposed to the idea, procedure, system, method of operation, concept, principle or mere data that underpins it. In the case of content generated by Gen AI, this is compounded by the fact that the way in which that content is expressed is largely, though not always, determined by the AI. Thus, copyright protection will depend on the extent to which a human author determines the expression of AI-generated output. Accordingly, where Gen AI simply assists a human author, such that the author retains and exerts sufficient influence over the final output, the resulting work may be copyrightable. Where, however, the content is entirely created by AI with trivial or negligible human involvement, then copyright protection will not lie. Similarly, a mere user prompt would not render the user an author: a prompt being an idea that has no material influence on the way in which generated content is expressed.
Interestingly, the US Copyright Office in the second part of its Artificial Intelligence Report took the same view: “…the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements. This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts. The Office confirms that the use of AI to assist in the process of creation or the inclusion of AI-generated material in a larger human-generated work does not bar copyrightability.”
Similarly, the EU Commission in its report on AI notes that: “Even if the connection between the human intervention and the AI-assisted output is increasingly remote, at this point in time, it is hard to conceive of content that is generated through AI that involves no human agency whatsoever. What is problematic today and for the immediate future is whether, and to what extent, a natural person’s involvement with the AI-assisted output – however remote – is sufficient for it to qualify as an intellectual creation.”
Computer-Generated Works
From a comparative standpoint, British law provides for and protects computer-generated works. Specifically, under section 9(3) of the Copyright, Designs and Patents Act 1988 (as amended), in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author is taken to be: “the person by whom the arrangements necessary for the creation of the work are undertaken.” This provision is silent on the question of ‘originality’. It has been said that whilst the provision “clarifie[s] that creations generated by a computer could be classified as works, [it] said nothing about how the originality of such works is to be determined. The particular problem that arises with computer-generated works is that it is difficult to see how the existing criterion of originality, which focuses on the relationship between the author and the work, can be applied to computer-generated works, which by definition, have no readily identifiable author.”
It remains to be seen whether the British courts read in an analogous standard of originality or treat computer-generated works as giving rise to a species of related rights. The Sri Lankan IP Act, however, has no equivalent provision.
Comparative Case Law
The following international cases are of interest and relevance:
Conclusion
It is clear therefore that Sri Lanka’s copyright law regime sits uncomfortably with relatively authorless content produced by Gen AI, and as a result, the IP Act may need future reform depending on how the technology continues to evolve. Gen AI, which can create high-quality content with minimal human intervention chafes against a system of copyright protection where the protection of authorial works is premised on the existence of a readily identifiable human author. The key takeaways are that firstly, originality and authorship are the main obstacles to the copyrightability of AI-generated content. Secondly, absent any further legislative reform, any claim of copyright protection over AI-generated output will be highly fact specific and depend on the extent to and the sufficiency with which human inputs determine the expression of AI-generated output.
Hashendra Jayasuriya
Legal Assistant
Drushika Amirthanayagam
Senior Litigator