There is a need to understand how Artificial Intelligence (AI) can be adopted into the IP ecosystem because it is the future of copyright, potentially. Globally, IP regulators are convening sessions to consider the nature and effect of AI in creating smart/auto copyright, amongst other IP rights. AI is an intelligence system that can receive data and interpret it to give a specific desired outcome, thus a simulation of human minds by programmed machines (Delipetrev et al. 2020). In this sense, the machines can reason, learn and analyze activities.
AI is conceived and increasingly understood as playing an active and accelerated function in the processes and roles associated with the creation of intellectual property the world over. The AI products used in the generation of IP, are typical works of programmers and developers amongst other players who are end users of AI-driven products (Kop 2019).
Therefore, the AI products’ users retain claims in copyright over the partially or wholly AI-generated work on the basis that the commands input through the AI systems yield AI works/copyright as an output. The human factor in produced copyright is an integral part of recognizing the primary author to benefit from the AI-led copyright.
Artificial Intelligence, as an enabler of innovation in day-to-day operations, is gaining momentum in Kenya as the country’s ICT landscape continues to hold conversations on the effect of the fourth Industrial Revolution (4IR), judging by increased technological advancements and increased automation of most processes and actions.
AI in this context has been interpreted as a simulation of human minds by machines, programmed to adapt to characteristics of conventionally human actions to generate art – music, sounds, gaming, ads etc. (Delipetrev et al. 2020). These machines or computer systems are programmed by feeding them specific information whose result is aimed at giving specific output.
AI in hindsight and context
In the 1960s, Professor McCarthy wrote about AI and referred to it as a ‘program with common sense’ because AI is programmed so that it can actively receive commands and act upon the said commands. AI exists in the simplest forms, such as Amazon’s Alexa, Apple’s Siri and Google Assistant. In more complex scenarios, it entails robots and other programs that have been tailored to respond intelligently to commands.
AI is unprecedented but its evolution was anticipated, even though its advancement has now been actualized faster than estimated. AI as an innovation of the mind raises questions of Intellectual Property Rights. Programmers who design AI products seek certain Intellectual Property Rights Protections. These innovations may require patent protection for the software and programs that actualize the use of AI.
Increased technological advancements from Western States are attributed to sufficient resources and capacity to accommodate these technological advances that have been backed by effective legislative intervention. Similarly, African States have continuously recognized the use and benefits of AI, and hence they are coming up with legislation and guidelines seeking to recognize and protect the rights that come with the use of AI.
The approaches to addressing AI issues in Africa include; the development of National AI Strategies, the establishment of AI Task Forces and Commissions, developing and adopting of AI regulations, and creating of awareness on the use and importance of AI. For instance, countries like Kenya, Uganda, Mauritius, Tunisia and Egypt have established various task forces and commissions to deal with and give recommendations on issues concerning AI.
AI affects other IPRs, such as patents (innovations for scientific solutions to problems). AI inventions may receive patent protection if they meet the conditions e.g. under Sections 21 and 22 of the Industrial Property Act in Kenya. The invention must be new, entail an inventive step, and be industrially applicable.
AI further affects Trademarks (marks, logos or designs) used to identify and distinguish products. For instance, when doing online shopping, algorithms are tailored to learn preferences and based on that make recommendations. This may result in a bias on the part of other brands as the algorithm may never recommend them to most persons, resulting in a drop in sales.
Further, AI is programmed to perform specific acts and it may not be able to distinguish original products from counterfeit ones and this is likely to have a negative economic impact on the product manufacturers as the consumption of counterfeit products will be on the rise.
Regulation of AI in Kenya
Kenya has taken tremendous steps towards embracing and recognizing AI. In 2018 the Cabinet Secretary of ICT, vide Gazette Notice No. 2095, established the Distributed Ledgers Technology and Artificial Intelligence Task Force. Further, in 2019, the Parliament enacted the Data Protection Act which is, a prima facie progressive legislation as it recognizes various challenges that arise with the use of technology.
AI has the potential of raising a country’s GDP as it gives the country a competitive advantage against other states and provides numerous investment opportunities for people who have ventured into the field of AI. Further, States have developed a keen interest in AI developments as they raise pertinent legal issues that require the legislators to look into.
Technology and its use require a level of regulation, failure of which challenges and questions are bound to arise. With technological advancements come new challenges that have made most States more inclined to understand AI and how they can incorporate it in an advantageous way.
In Kenya, the Blockchain and AI Taskforce in 2018 proposed the adoption of legislation (Ministry of Information, Communications and Technology 2019). The proposal was informed by the fact that AI inventions are developments of technology, and there are questions as to the ownership of the intellectual property that arises from AI creations. Therefore, if protection is not afforded to these creations, most people will likely be discouraged. This will put Kenya at a disadvantage as more tech-savvy investors will be disinclined to invest in a country where they are not accorded the requisite protections.
Practical application of AI in various forms
AI works in a way that when developing a program, the Programmer curates a specific code to run the programme. The code comprises the algorithm (steps followed to achieve the desired results). The algorithm is structured through machine learning, neural network, deep learning, computer vision, natural language processing, the Internet of Things, and graphical processing units.
Machine learning works by feeding specific data into a computer system to identify specific patterns and make decisions based on them, thus a creature of habit. A good example is autonomous/self-driving vehicles, programmed to perceive and avoid obstacles. In its advanced form, machine learning is referred to as deep learning, which deals with larger amounts of data, enhancing efficiency. It manifests as a clown of the human brain, more accurate than the human brain. It ensures total automation and no human intervention is necessary, e.g. Apple’s Siri and Amazon’s Alexa.
Neural networks, on the other hand, possess problem-solving abilities and are programmed to recognize hidden patterns, analyze and classify data. This has been important, especially in the medical field (diagnosis).
The Internet of Things (IoT) enables machines to learn a person’s preferences and recommend best suited products for that person. It sparks issues of data protection. Web-enabled smart devices, with sensors and communication hardware, data may be collected from one device to another with or without human intervention, e.g. smart homes where temperatures can be controlled automatically.
AI works in the legal profession, for example, in contract review, case outcome prediction and apportionment of liability. Analysis of programmed data enables review of contracts, with the desired result being flagging mistakes to achieve an ideal contract. In the same way, AI can be adopted in the generation of works capable of copyright protection, as discussed below.
Comparative cases on AI associated copyright
The 2022 Colorado State Fair Art Competition presented a chance for Jason M. Allen (Pueblo West, Colo) to win the blue ribbon for the emerging digital artists category. His work, ‘Théâtre D’opéra Spatial’ was the first AI-generated work to win an award at the fair. Other artists claimed that he did not come up with the art as it was AI-generated through ‘Midjourney’.
‘Midjourney’ has an ‘imagine’ command which the artist selects and keys in the text and using the ‘imagine’ command, the application responds by turning the text into hyper-realistic images. Can such art receive copyright protection? If all an artist has to do is key in texts and the image is created, is there an element of work involved?
‘Midjourney’ the bot, has been exposed to various art of other artists, which raises the question of originality. Training AI systems is like a painter finding inspiration from other artists’ work; hence the resultant issue is where the work generated by AI is substantially similar to already existing work.
Originality can be resolved by answering the question of whether similar work existed or not, absent of which the machines would not generate art on their own. As long as this question remains unanswered, the originality of AI-generated art is questioned.
AI-generated work can be eligible for copyright protection on the premise that AI has been used to complete paintings by very famous artists, e.g. the Rembrandt. AI gave people a visual of what the complete Rembrandt would look like. AI has the intelligence to predict how a human would have completed a painting.
In my opinion, where there already exist similar works to be imitated by AI, the smart product of AI should not be afforded copyright protection. There is a delicate balance between AI’s autonomously generated work and work generated by a human author/artist with the help of AI.
The US Copyright Office recently considered whether AI-generated art could receive copyright protection. Steven Thaler made his first application for registration of the work ‘A Recent Entrance to Paradise’. This was on behalf of the ‘Creativity Machine’, an algorithm programmed to autonomously generate the art. The first application, lodged on 3rd November 2018, was rejected (lacking the human authorship aspect) (Long, 2021).
The Applicant applied for the Office to reconsider the rejection arguing that the requirement for human authorship was unconstitutional. The Office upheld the rejection, proposing that even the courts in the US require a mandatory nexus between the human mind and a creative expression for copyright protection. However, the courts have been quiet on the extent of AI involvement required for the work to be copyrightable.
Cases of robot-made art may further raise concerns on IP protection. For instance, Ai-da, a robot created by Aidan Meller, and identifies with the pronouns she/her was among the first bots to actually paint art. The bot, whose creation began in 2019 and was completed in 2022, has been programmed to use a brush to create paintings. It makes decisions by interrogating various aspects to develop an original and creative art. The question that then arises is not whether robots can create art but rather does this amount to work capable of being copyrighted?
The gaps in AI copyright recognition
The AI taskforce report on AI in Kenya illustrates the gaps in developments brought about by use of AI in Kenya. To address the gaps, it is critical for a governing framework to be developed, including putting in place regulations and legal controls on the use and application of AI, especially in relation to intellectual property.
Intellectual Property Rights seek to encourage creativity. However, if technological developments cannot be protected, people will likely be discouraged from being innovative.
The use of AI by creatives, authors and artists has led to many moral, ethical and legal questions especially when seeking Intellectual Property Protection. Among the legal questions are whether a program developed should receive patent protection, whether work developed by AI can receive copyright protection, whether the use of AI results in IPRs infringement in certain cases, and whether the AI can claim infringement of copyright.
One major concern that relates to copyright proprietorship in the AI-generated work, whether it is art, books or any other copyrightable work. For instance, in AI generated art there is the Programmer who develops the algorithm for the program and based on the algorithm the machine can receive commands from the user and respond by giving the desired outcome. The challenge is determining the owner of such work and who can freely exercise all rights conferred by copyright. However, we recognize that the Programmer may copyright their program but the extent of this protection is debatable.
As a country, Kenya has always kept track of technological developments to avoid lacunas in policy and legislation. However, technological advancement occurs faster than people anticipate, even before the legislators can come up with policy and legislative interventions. For instance, the Copyright laws in Kenya cover literary works, musical and artistic work, audio-visual work and broadcasts. Section 22 of the Copyright Act provides for the works eligible for copyright protection by setting out the requirements for protection of literary, musical or artistic work which are: sufficient effort be expended in the making of the work to make it an original creation; and be reduced in material form.
The current Copyright laws did not anticipate technological advances and their impact on Copyright laws. For instance, Section 2 of the Copyright Act 2001 defines an author as a person. The legal definition of person covers natural persons, and extends to a legal person like a company. The resultant question is what happens to ‘Ai-da’ the robot that can paint using a painting brush just like an artist. Do we then recognize her as a natural person or a legal person? There are numerous gray areas, and the lack of concise laws does not aid the situation.
Our legal framework does not anticipate a scenario where an author could be a machine i.e. generation of artistic works by advanced technologies such as AI without substantial human effort or intervention. For example, the AI art Maker or the Dall-E-2 Applications require a person to give a description, and based on that, the applications will create artistic output.
Considering the provisions of the Copyright Act, it does not specify that an author must be a natural person but clarifies that an author is the person who first makes or creates artistic work. It is not clear whether the author will be the machine as it was the first creator of the work or the person who gave a description for the work to be created. Further, the law does not expressly provide protection for the description that is keyed into the application, however creative it is.
The resulting question is whether it is important to accord the requisite protection for such description. However, if we are to rely on the already existing Copyright Laws, then the description must be expressed in material form for it to qualify for any intellectual property protection.
Copyright laws in Kenya have no express provisions on whether machines can receive authorship recognition, judging by the definition of ‘author’ under Section 2 of the Copyright Act. Section 22 of the Copyright Act merely emphasizes that work shall only receive copyright protection if sufficient effort has been expended on making the work, to give it original character.
There is no clear policy direction or legislative provision defining what amounts to ‘effort’ as defined under the Copyright Act. However, decided cases demonstrate that authorship requires to be attributed to a human mind. This does not, however, mean that attribution cannot extend to programmers and AI inventors and creators. There is no suit yet in Kenya on the AI Copyright to test the theory that a human mind has to be purely a natural person.
Case scenario is where a person keys in data into a machine and in response to the command, the machine then creates a piece of art. The law did not anticipate such developments, hence difficult to answer the question of whether there is sufficient effort expended to qualify auto-art for copyright protection.
The nexus between AI and art continues to pose ethical and moral issues. Conventionally, intellectual property rights extend to original works of the mind. Where AI is used to create art, the question of originality arises as an ethical/moral concern. The way AI works, the machine is given data and based on that data, it gives a desired output. For art, the machine is exposed to art from different artistes, learns patterns, and produces art depending on the demand. This raises the question of whether the art generated is original.
Section 2 of the Copyright Act 2010 defines artwork as an original work of visual art created by an artist or produced under their authority. Further, section 22(3)(a) states that copyrightable work is work that sufficient effort was expended to it giving it an original character. This takes us back to the question of what amounts to sufficient ‘effort’.
As highlighted, ‘Midjourney’ works by being exposed to numerous paintings, making it possible for it to emulate the strokes of the brush just like other artists hence producing art. All this leads us to the same question of whether AI work is original work. Further, whether by virtue of the machine being exposed/fed other artist’s work then, it takes away all the originality from the work created by AI.
Potentially, works produced by AI could face plausible claims of copyright infringement since the results are based on already existing art. One of the major steps towards resolving the question of originality is determining whether AI-generated art would still be possible without the machine being exposed to art by other artists.
Additionally, when using AI to generate art, the question of who the claim will be made against arises in the event of an infringement. Whether the claim will be against the Programmer, the user or the AI machine, where it is owned through a registered legal entity. This difficulty arises because the author or artist has not been clearly established. The non-existence of a clear framework of ownership also raises issues regarding who gets the benefits derived from the AI-created art.
Traditionally, art is an expression of emotions and feelings. Consequently, the use of AI takes away the emotional aspect of art and creates mostly superficial art. Art lovers then find it difficult to connect with the art compared to human-made art. For instance, Ai-da the robot, was interviewed and it stated that it does not have emotions but it can draw what it sees. At the end of it all, art will be created, but whether that art connects with the people who consume it. Another question that arises is whether we recognize the robot as a person and that the art it has created can be fully attributed to it since it was created and even has a name and gender.
Conclusion
AI is not an entirely new concept, and these continuous technological advances popularize its use amidst challenges. The major challenge is the lack of policy direction and a robust legislative framework to address questions that arise. There is a need for stakeholders in the ICT Sector, including legislators, to consider how AI-related Copyright can be regulated in terms of recognition and attribution of authorship.
The Regulation should be all-spectrum, including sector players such as the Law Society of Kenya, considering that AI is taking shape, even in the legal sector, where contract review algorithms and other legal solutions providers are becoming a norm. Various stakeholders and key players in the ICT and intellectual property realms should be at the forefront, ensuring Kenya takes sufficient measures to keep us a step ahead in embracing various technological advancements.
It is noteworthy that Kenya has taken formative steps to address the use of AI, such as through the AI task force and the Kenya Copyright Board (KECOBO). KECOBO’s Issue number 38 provided a platform for the conversation about Artificial Intelligence and Copyright, where one of its employees provided insights, identifying concerns about AI and IPRs. This goes a long way in achieving the end goal of sufficiently regulating the use of AI to ensure the protection of creatives. However, merely putting the conversation out there is not sufficient to achieve the end goal, as decisive action needs to be taken.
In my view, Kenya’s current intellectual property landscape is designed in a way that does not anticipate rapid and disruptive technological advancements. AI innovations could shape how intellectual property rights are conceived and protected. The import of AI in the copyright sphere must be recognized. This will lead to legal protections, limitations and regulation in light of the developments in technology, rather than leaving the area uncharted.
This paper holds the view that notwithstanding the challenges, the current legal framework on copyright should remain the same, with necessary adjustments to incorporate AI and other technologies. In the alternative, the AI aspect of Intellectual Property should have its own sui generis framework for protection. Legislators should make laws specific to AI-generated copyright works. For instance, current copyright protection lasts 50 years. In light of the disruptive nature of AI, there is perhaps a need to consider protection for shorter timelines, e.g., 5 years for exclusive ownership, to allow for innovation and competition amongst users of AI systems.
The works can thereafter be available for renewal of protection or made available for use by other people because programmers and developers keep advancing and upgrading AI-driven algorithms, making it possible to come up with numerous works continuously. This increases the ability to produce numerous works within a shorter time than a human being.
Failure to regulate AI in the Intellectual Property realm may demotivate creatives for lack of incentives to create works that cannot be protected from infringement and give them exclusive rights. Ultimately, the recurring questions are: Can AI AI-generated works receive full copyright protection? What is the future copyright landscape looking like in light of AI-driven technologies in the creative industry?
We can only wait in anticipation.
References
Delipetrev, B., Tsinaraki, C., & Kostic, U. (2020). Historical Evolution of Artificial Intelligence.
Long, R. (2021). “AI Creations: Legally Protected”. https://cyberlaw.stanford.edu/blog/2021/04/ai-creations-legally-protected
Kop, M. (2019). AI & intellectual property: Towards an articulated public domain. Tex. Intell. Prop. LJ, 28, 297.
