AI and creativity – Are the outputs of machines protected by copyright?

 But are AI-generated outputs protected by copyright and who should be considered the author of these works? Moreover, who should bear the responsibility if AI infringes copyrights held by others when creating its own works?

Over the past few years, the development of various AI technologies has progressed at a rapid pace due to, among other things, the increase in computing power and the development of new algorithms. Artificial intelligence generally refers to a machine that has characteristics of human intelligence and is able to independently perform a given task. In creative industries, artificial intelligence is usually trained to perform a task using machine learning techniques.

A great example of machine creativity is the music-composing robot AIVA (Artificial Intelligence Visual Artist), which is used to create classical music. AIVA has been accepted as a member of the French collection society, SACEM, and it produces new types of songs using existing classical music as training data [1]. However, the creativity of machines also induces a number of new and open legal questions.

Are the outputs of artificial intelligence protected by copyright?

The requirements for copyright protection of AI-generated outputs are the same as in other situations. This means that the final output must be considered sufficiently independent and original. However, AI applications usually work in limited ways, and their creativity is often based on training data collected and fed to them by humans. Therefore, it is currently unclear whether the works produced this way may be considered sufficiently independent and original in a way that gives rise to copyright protection. It is possible that the more independent the creative activity of artificial intelligence is, the more probable copyright protection becomes.

If the requirements for copyright protection are met, we still need to solve another rather complex dilemma, that is, discerning the author of these AI-generated outputs. This question is especially challenging in cases where the outputs are created with little to no human intervention. Would it be possible to grant the status of author to the AI application itself?

At present, this rather futuristic question must be answered in the negative. Even though the Finnish Copyright Act does not define the concept of author, the act still implies that only a natural person may be considered the owner of copyright. This is also implied in the Infopaq ruling issued by the European Court of Justice. The Court stated that the originality of a work requires that the work be a reflection of the author’s creativity. In other words, the work must reflect the personality of the natural person who created it.

Granting copyright to a machine would also be against the fundamental understanding of which entities in general have legal rights and obligations in our legal system. According to the current understanding, machines and other artificial agents are not accorded such legal personality.

In light of the above, the copyright of AI-generated outputs must be assigned to a person (or a group of persons) who have contributed to the operation of the application. In this case, the AI application would simply be regarded as a tool similar to an artist’s paintbrush. However, as AI technologies further evolve, this assessment is likely to become even more multi-faceted.

Who is responsible for damage caused by AI-generated works?

Since the emergence of early AI technologies, there has been much concern over autonomous machines taking over and causing uncontrollable damage to humans. However, it must be noted that the risks posed by AI greatly vary between different industries. The applications currently used by creative industries, such as painting robots, are evidently not as high-risk as those used in healthcare or transport. Nonetheless, there may still be certain risks involved, for instance, the risk of infringing another person’s copyright.

The risk of copyright infringement is involved in both the creative process and the final outputs. Similar to an artist who is constantly influenced by his or her environment, AI applications also use a vast amount of data to generate creative outputs. This data may be protected by copyright or related rights, in which case their unauthorized use may construe a copyright infringement. Furthermore, the final output of the creative process may itself be infringing.

If the infringement causes compensable damage, the question as to who is liable for such damage arises. Is it possible to hold the AI machine itself liable if the infringement is characterized by the absence of substantial human intervention? As stated above, AI applications are currently not considered legal persons in the Finnish legal system. Therefore, the above question must again be answered in the negative. The responsible party must hence be sought among the persons who contributed to the operation of the machine. It may be possible to bring a claim for compensation against the artist who utilized the AI. If, on the other hand, the copyright infringement can be attributed to a manufacturing defect in the application, it may also be possible to claim compensation directly from the manufacturer.

Conclusion

As outlined above, the copyright questions associated with works created by AI may be complex. The challenges posed by AI are likely to increase in the future as the role of the software developer or user further diminishes. Indeed, issues related to AI and copyright have been much debated at both national and international levels. However, it remains to be seen whether legislative measures to adapt to these advances will be taken any time soon. 

Authored by: Sofia Wang