ChatGPT: The Pandora’s Box Opened?

As for the Malaysian perspective, while the passing of an AI Act may not seem imminent, perhaps as a first step, the provisions under the CA should be clarified in order to curtail the impending effects of ChatGPT

Since the 1980s, the world has been exposed to the notion of artificial intelligence (“AI”) through blockbuster movies such as The Terminator and Robocop, albeit in a fictional account. In reality, the defeat of World Chess Champion, Garry Kasparov by IBM’s supercomputer, Deep Blue, in 1997 raised the possibility of AI being superior to human intelligence. In fact, AI has been the subject of debate for decades, either from the ethical or legal aspect. The launching of CHAT Generative Pre-trained Transformer (“ChatGPT”) by Open AI in November 2022 shocked the world when an AI chatbot, which was able to, among other things, draft legal documents, essays, poetry, compose songs, translate documents, generate images, solve mathematical problems, was introduced. However, when issues such as plagiarism and infringement of intellectual property rights started to arise, one was left wondering whether the introduction of ChatGPT had in fact stirred up a storm. In particular, copyright owners, when faced with the possibility of ChatGPT reproducing their copyrighted works without their consent, looked to the law for a legal framework sufficient to address their concerns.

Chatgpt Vs The Current Landscape of Copyright Laws in Malaysia

The uncertainties and lacunae in the Copyright Act 1987 (“CA”) as set out below render the current legal framework in Malaysia uncertain as to its sufficiency to combat the aftermaths of ChatGPT pertaining to copyright ownership or infringement issues: -

A. The identification of the author for works generated by ChatGPT

The determination of authorship of works generated by ChatGPT is probably one of the most unclear issues under the current copyright regime. This issue often relates to whether “human authorship” is required before ChatGPT can claim authorship over the works that it generates. Pursuant to the language of the provisions of the CA, this would depend on the type of works involved owing to the following reasons: -

a) Based on the definition of “author” under section 3 of the CA¹, the author of literary, musical and artistic works would be the writer or the maker, the composer, or the artist of the respective works. This suggests that “human authorship” is required for these works. On the other hand, the definition of “author” for the other types of works such as photographs, films, sound recording, or broadcast would be the person who undertakes the arrangements to make these works, which arguably may include a natural person or a body corporate.

b) The computation of the duration of copyright fortifies the abovementioned interpretation. Section 17 of the CA provides that for literary, musical or artistic works, the duration of copyright is computed based on the life of the author plus 50 years after his death. On the other hand, for other works such as published editions, sound recordings, broadcasts, films and works of Government organisations and international bodies, the duration of copyright is computed based on the act of publication or making of the work, as the case may be.²

That being said, the abovementioned interpretation was muddled by the decision of the High Court case of Creative Purpose Sdn Bhd & Anor v Integrated Trans Corp Sdn Bhd & Ors³ where it was held that: -

“if an ‘author’ can also mean a ‘qualified person’ and ‘qualified person’ is defined by the Act as including a body corporate⁴, it cannot be right to read the word ‘author’ in s 26(4) as referring only to natural persons.”

This decision arguably creates the possibility for the recognition of authorship regarding works generated by ChatGPT provided that the name of the individual(s) or a corporate body is displayed on ChatGPT’s system. With respect, such decision does not sit well with the language and spirit of the CA. Firstly, this decision is contrary with the definition of “author” under section 3 of the CA as well as the provisions in relation to the computation of the duration of copyright work. If a body corporate is qualified as an “author” it would result in the said body corporate being able to“claim a perpetual monopoly over their works to maximise the economic returns of copyright protection.”⁵ The Judge’s reliance on section 26(4) of the CA⁶ is also arguably misplaced as this section relates to the presumption of ownership of copyright, which is a distinct concept from “authorship” of copyrighted works. Perhaps a reconciliation between the definitions of “authorship” and “qualified person” is required to resolve this issue.

B. The absence of the definition of “computer generated work”

Section 7(1) of the CA expressly sets out the categories of work which are eligible for copyright protection. This includes literary works, musical works, artistic works, films, sound recordings and broadcasts. Computer generated works (which includes works generated by ChatGPT) are not featured in the definitions for these categories of works. It is thus doubtful whether works generated by ChatGPT are eligible for copyright protection.

C. The possibility of Open AI being liable for copyright infringement

In light of the widespread usage of ChatGPT, it may be impossible for copyright owners to enforce their rights in their copyrighted work effectively against all direct infringers. The only practical alternative would be to go against Open AI, the developer of ChatGPT. Section 36(1) of the CA expressly provides that “copyright is infringed by any person who does, or causes any other person to do, without the licence of the owner of the copyright, an act the doing of which is controlled by copyright under this Act”.While the phrase “…causes any other person to do…” somehow suggests the possibility of OpenAI being liable for copyright infringement, this phrase has yet to be interpreted by the Malaysian courts. Case laws in the UK suggests that if the element of control on the part of a defendant can be shown, it can be deduced that the defendant has caused third parties to commit copyright infringing activities.⁷ A similar interpretation of the phrase may be adopted by the Malaysian courts. That being said, even if this interpretation is applied, the factual issue would be the extent of control exercised by OpenAI which in practice may not be easy to determine.

An Artificial Intelligence Act to Safeguard the Interests of Copyright Owners?

In light of the lacunae as set out above, one may wonder whether there is a need to have a legislation to regulate AI generated works. On 11 May 2023, a key committee of lawmakers in the European Parliament approved the European AI Act (“the AI Act”).⁸ Essentially, the AI Act seeks to regulate AI where the AI system will be categorized into 4 levels of risks: unacceptable risk, high risk, limited risk and minimal or no risk. Thereafter, requirements and obligations (among others, the need to measure, assess and mitigate risks to fundamental rights) will be imposed on an AI system based on the level of risks that it is categorised. While the AI Act raises some concerns amongst the AI players and the tech industry, there are no specific provisions to address the IP issues that could arise. As for the Malaysian perspective, while the passing of an AI Act may not seem imminent, perhaps as a first step, the provisions under the CA should be clarified in order to curtail the impending effects of ChatGPT.

Disclaimer – The views expressed in this article are the personal views of the authors and are purely informative in nature.