Artificial Intelligence: challenges for the existing copyright framework

The common perception of artificial intelligence (AI) is that of SkyNet in the Terminator series, the hosts in Westworld or Ava from Ex Machina; unnerving and implausible in equal measure. The application of AI is generally represented through the image of computers that can communicate, behave and reason in the same way as a humans. With this in mind, many would dismiss AI as mere science fiction; a concept that we would not want to encourage in any meaningful way.

However, artificial intelligence and the application of AI in modern life is more wide-ranging than many would anticipate. In broad terms, AI is a collective expression for a myriad of disciplines, from machine learning systems, in which computers are given the ability to learn and improve with experience, to neural networks, as a form of machine learning loosely based on the human brain. The common principle that unites each of these disciplines is the ability to replicate an aspect of human cognition, such voice recognition available in mobile devices or chatbots on websites.

Given the advancement of AI over recent years, combined with the increasing potential for a computer to be able to create things autonomously without human intervention, it is inevitable that challenges will be posed of the existing legal framework, particularly in relation to copyright and other intellectual property matters.

In this article, we consider how AI systems and content created by AI systems, is treated under the current copyright regime in the UK and, most importantly, if that regime is fit to accommodate AI technology moving forwards.

Ownership

Under the Copyrights, Designs and Patents Act 1988 (“CDPA”), the legislation governing copyright in the UK, copyright can subsist in various categories of works, including literary, dramatic, musical or artistic works. Crucially, for the purposes of this article, the definition of a “literary work” includes computer programs and databases. As such, it is commonly accepted that the primary legal protection for software is copyright.

The owner of the copyright in a work is generally the person who creates the work, referred to under the CDPA as the “author”. Traditionally, software is developed by a human author through writing source code and therefore it is clear that the owner of the copyright in the software is the person who developed it. In broad terms, if that person developed the software in the course of their employment, the employer will be the first owner of the copyright (subject to any agreement to the contrary).

However, in considering the ownership in AI systems, or copyright works that are created autonomously by an AI system, the position regarding authorship and ownership of copyright becomes distorted. The underlying logic of many AI systems is developed by the system itself, as a result of undergoing a training process in which the system develops its own decision-making algorithms and rules by practicing and improving using large volumes of training data.

In these circumstances, much will depend on the level of human intervention and input. If a human has developed the AI system and is involved in implementing the logic, then it is likely that the system and its output will belong to that human author. On the other hand, it is less apparent who owns copyright in a fully autonomous AI system that has, through its training process, developed its own logic.

The CDPA does accommodate the concept of “computer-generated works”, under which if there is no immediate human author, the author is the person by whom the arrangements necessary for the creation of the work are undertaken. As such, in straightforward cases, it may be the organisation that controls or directs the use of IT systems that will own the logic and output of AI systems.

However, in view of the collaborative nature of AI systems, it is possible that often one party may develop the AI system’s framework, whilst a separate party will undergo the training process to create the specific logic. For example, a supplier of an AI-based platform may licence the system to a customer, whilst the customer will implement its own data to “train” the system, and develop the underlying logic and decision-making processes. In these circumstances, it is not clear who has made the arrangements necessary for the creation of the work. Is it the supplier of the AI system, who has provided the platform upon which to produce the output, or is it the customer who has undertaken the training process, but has not developed the underlying system? There may also be circumstances in which the AI logic is developed by large teams of programmers, with multiple parties involved in designing algorithms and determining data sets to be analysed. Again, the matter of who owns the copyright is far from clear.

In view of this, the position under the current regime could increasingly give rise to disputes over the ownership of the AI logic, which in many cases may have significant value. In the absence of certainty under the relevant legislation, ownership will largely be determined by the terms of any contractual relationship between the parties. As such, it will be important to ensure that on any collaborative AI-based project, the provisions dealing with ownership of any intellectual property are clearly defined and understood by the parties involved.

Originality

The CDPA also provides that, for a work to be capable of benefitting from copyright protection, it must be an original work. Originality is defined by reference to the work being created through the author’s own skill, judgment and individual effort, and cannot be copied from other works. Whilst the threshold for establishing originality is generally low (and, with the exception of works of artistic craftsmanship, the work does not need to have any artistic merit), there is some doubt over whether certain works developed by a trained AI system would be able to meet the originality test.

In many cases, the output of an AI system will not have involved the author’s own skill, judgment and individual effort, but rather, will be a product of the algorithm or training process undertaken by the AI system. Although a human may have been involved in determining the arrangement necessary for the creation of the work, the nexus between the arrangement and the actual output may be difficult to establish. Likewise, as an AI system cannot be an “author” under the CDPA, it cannot not exercise its own skill, judgment and effort in developing the work to meet the originality criteria. The position under EU law is similar, insofar that the work must be the “author’s own intellectual creation”, which again assumes that a human author will always be identifiable.

Given that the CDPA was implemented 30 years ago, when AI may have still been considered far-fetched, it is understandable that the test for originality does not entirely anticipate the complex challenges posed by the implementation of AI systems. However, it gives rise to some potential concerns, not least that works created through autonomous AI systems (with potentially significant value) may not be protected by copyright, and therefore can be copied by others without recourse.

Conclusion

On any analysis, it is clear that to some extent the copyright regime in the UK will need to be addressed as a result of the likely influx of AI-based technology. Given that copyright is the main legal protection available in relation to software, issues of ownership and originality are fundamental matters that are the cornerstone of the copyright framework, and need to be sufficiently well-defined. Moreover, there are also other elements of copyright, such as copyright infringement, that will be need to be considered in the context of AI systems.

In the recent paper published by the Select Committee on Artificial Intelligence, the committee pushed for clarity on these types of issue and for the adequacy of existing legislation to be considered further. On that basis, it is entirely possible that the UK government will consider new legislation to make the copyright regime, and more widely, other intellectual property rights, more AI-friendly. In the meantime, organisations will need to ensure that if they are involved in any AI-related projects, that the contractual relationship between the parties is clear and precise on these otherwise uncertain issues.

Note: the content of this article is for general information only and does not constitute legal advice. Specific legal advice should be taken in any specific circumstance.

If you have any questions on any of the points covered in this article, please do get in touch with our Corporate & Commercial Team on 01603 610911.