Nowadays, virtually any part of society has been improved by technologies that involve artificial intelligence (AI). In its basis, a work of art is a direct creation of a living person who then claims ownership over the content. However, materials that are subject to copyright laws can also be generated by AI, making many past law practices obsolete and requiring to be updated in order to accommodate new technological advancements. It is vital to assess the complexity of mechanisms behind AI that create such content, as it should be possible to conclude if the input by their creators should suffice the requirements of originality. I would like to assess the ways AI interacts with copyright law through examples of cases from several countries to determine the possibility of its application.
The way people operate information continues to change at a rapid pace ever since humanity has entered the digital era. Models that simulate writing, painting, and composing are already freely available to users online, while more advanced programs are being actively developed by high-tech studios. However, their output does not fit into the modern understanding of a unique work since the program will generate the same outcome if given the exact same initial data, putting the uniqueness of such art into question.
Before I discuss the possibility of integrating AI into copyright laws, I would like to review the existing legal grounds for this change. For example, The United States already has a history of complications with copyright laws related to content generated by non-humans. The United States Copyright Office has refused to give rights to a person whose camera has been used by a monkey to take a selfie, after which the country’s copyright practices compendium has been expanded with a statement that an artwork must be generated by a human being in order to be eligible for copyright protection. On this critical note, the current policies do not allow the claims of ownership on artwork not produced by a person directly.
Some of the European Union countries’ laws are more lenient towards AIs and their works. The UK Copyright Designs and Patents Act 1988 allows considering a person without whose creative input an artwork would be impossible to create as an author in copyright claims. At the same time, traditional laws often require an object of copyright to reflect an author’s personality, which AIs apparently lack. At best, AI-generated artworks can be deemed as a trade secret for protection until a better option appears.
In turn, Australia presents a challenge for copyrighting AI-generated content. Undoubtedly, any literary work or even a written program, including AI itself, falls under its copyright law as outlined in the Copyright Regulations 1969 Act. However, lawyers have found that there is an explicit requirement that copyright can only be acquired by a person, such as an Australian citizen or resident, qualified to be its author. Moreover, the 2009 case of IceTV Pty Ltd v Nine Network Australia Pty Ltd gave people the direct confirmation of this fact: there must be a significant artistic effort for a work to be considered copyrighted by a person claiming its rights. It is challenging to state that the actions of AI can fit into this context, making any claims of copyright ownership generated by a machine highly dubious in Australia.
The debates around the issue have caused organizations to begin assessing the position of AI-generated content in law. In 2020, the World Intellectual Property Organization held a conference where a set of recommendations to governments across the globe was proposed with an intention to change the very concept of being an author. Copyright policymakers have discussed what effort should be counted as sufficient for a work to be considered entirely original. At this time, no consensus has been reached, and no adjustments have been made to common copyright laws, yet the changes are bound to occur in the following years.
There are three major barriers to this change that remain largely unaddressed by policymakers. First of all, the very definition of a person in legal documents cannot fit a machine-executed algorithm. Yet another slew of issues is guaranteed to arise from this change, as copyright protection often has a length defined by years after the author’s death. A problem in establishing ownership also comes from the fact that different people create and train AIs. Which person deserves to be considered as a rightful owner of generated content is not clear. Another critical issue is the difference between forgery and artwork generation. Currently, machines merely adopt the existing content provided to them by their operators, gradually giving structure to common elements within a data set. This batch of information is later utilized for producing a work that exhibits these elements, sometimes even pulling parts of them into its output. If countries are to recognize the copyright ownership of machine-produced art after its algorithms’ authors, there must be boundaries at which mere replication of the existing artworks will be considered forgery.
However, it rapidly becomes clear that it is possible to consider people who wrote and trained an AI that produced artwork to be its legal authors. There are degrees of training, and the complex algorithms behind each generated artwork gradually create a unique model out of the same data set. With sufficient input from a trainer or an AI author in the form of additional layers of human additions, it would be possible for courts to reconsider the necessity to add a touch of personality directly. There is also a need to redefine forgery in this context, although this issue can be resolved by assessing the extent of modifications performed by an algorithm on a data set. It becomes apparent that the outdated views on information in law frameworks cannot support the modern developments, ushering governments to accept the new reality of the digital era. I do believe that AI-generated art will not be distinguishable from works made by humans, at which point the lack of legal support will be a major issue for copyright agencies.
In conclusion, AI-generated content is rarely protected by copyright law, yet there are reasons to believe that there will be changes in common practice that will outline the ownership of this material. Many copyright laws require a touch of personality as a prerequisite for original content. It can be argued that AI’s creators put their character into works that will be produced by machines when writing their core and teaching them the foundations of creating art. However, as of today, very few countries present a chance for AI creators to get their works recognized as protected by copyright laws.