The field of artificial intelligence has grown rapidly over the past fifty years. Modern artificial intelligence software is able to produce work that has never been created by computers before. For example, artificial intelligence can be trained to produce art, music, movies, poetry and stories. Therefore, the question arises, what is the law when artificial intelligence copies a work protected by copyright? Who will be responsible for copyright infringement – the artificial intelligence software, its creator or the person using the product? Is there a need for a new legal system that will adapt the changing reality to the new technology of artificial intelligence or can the existing system be used with limited adaptations of the legal reality to the technological reality.
In this article we will explain what artificial intelligence is, what copyright is, and we will discuss the question of adapting the legal system to the legal challenges created by the new technology, in addition we will discuss the question of responsibility for copyright infringement in a work created by artificial intelligence.
What is artificial intelligence?
The field of artificial intelligence (Artificial intelligence – AI) is a branch of computer science, which focuses on computer operations that normally require human intelligence to perform them. Artificial intelligence is the computer’s ability to extract useful information from a data set and process it to produce results that match the response of an intelligent human. Generally, artificial intelligence includes abilities of learning, understanding languages, recognizing patterns, solving problems, responding to questions, trial and error.
Today, almost all artificial intelligence techniques are based on a technique called “machine learning”. Machine learning uses computer algorithms that can learn or improve performance over time on a specific task. These algorithms allow the AI software to understand the best way to perform the task without being programmed by specific instructions.
An example of software learning is teaching artificial intelligence software to recognize a picture of a dog. First, the artificial intelligence software is shown many pictures of dogs. The AI software program contains many nodes (neural network) that work together to focus on different aspects of each image. The learning process of the artificial intelligence software is very slow at first because the artificial intelligence does not receive instructions regarding the characteristics of the dog on which it should focus. However, after the AI software goes through thousands of images of dogs, it can learn the dog’s characteristics and can even create its own unique dog faces – similar to how a baby learns to recognize a dog [1].
What is copyright?
Copyright gives the owner of an original work the exclusive right to copy, publish, perform, broadcast, make available to the public and rent the work. Copyright protection applies to an original work that is a literary, artistic, dramatic or musical work, fixed in some way, as well as to a record.
According to the ruling in Israel, an “original” work is a work that originates from the creator (that is, a work that was not copied from another). In addition, there is a minimal requirement for personal and creative investment in the work. Copyright does not protect an idea but the original expression of the work.
What is copyright infringement?
A copyright is violated when someone performs one of the actions exclusively assigned to the copyright owner (copying, publishing, performing, broadcasting, etc.) without obtaining their permission. Since it is a proprietary right, the copyright owner is not required to prove intent or knowledge of the infringer. Also, the law imposes responsibility on those who commit an indirect copyright infringement (for example, trading in an infringing work or assisting in the infringement of copyright). The proof of responsibility for indirect infringement requires proof of knowledge on the part of the infringer. However, it is not required to prove actual knowledge, but closing your eyes is enough.
Copyright law establishes permits for making certain uses that allow the infringer to justify the infringement of the work. For example, under the “fair use” protection, the infringer can use the work for the purpose of self-study, research, criticism, review, journalistic reporting, making quotations, or teaching and examination by an educational institution.
Liability for copyright infringement by artificial intelligence software
Artificial intelligence software is usually used to create works for personal or industrial purposes. Works are eligible for copyright protection if they are original, with most definitions of originality requiring a human creator. In most jurisdictions throughout the world, it has been determined that only works created by humans can be protected by copyright. Therefore, it is not surprising that the infringement provisions in many copyright laws around the world also refer to infringement by a person.
When a person infringes a copyright, the creator can file a lawsuit against the infringer for copyright infringement. Since in most jurisdictions in the world artificial intelligence systems do not have legal personality, and as of today they cannot own assets, there does not seem to be any benefit in trying to file a lawsuit against them. If so, who should be held accountable when AI software infringes third-party copyrights?
One group is the creators of artificial intelligence – creating effective artificial intelligence software is a complex craft that often involves a large number of people or officials. Some people are responsible for writing the software code that underpins the AI, others will be responsible for “training” the AI software by feeding or referencing data sets that may contain copyrighted works.
Another group is the entities that request output from the artificial intelligence – the artificial intelligence creates works based on a request from an entity that requests the artificial intelligence to draw something for them.
It is assumed that each group has a different legal risk in relation to the question of copyright infringement.
Revolution or evolution?
Artificial intelligence is going to become an integral part of the business environment. This reality may pose new challenges to the law and justice system. Do the existing legal tools respond to business requirements in the changing technological environment? Or does the transition to an electronic environment require a reform of the existing legal arrangements?
At one extreme are various legal commentators who believe that a new framework of legal science is required to deal with this new and emerging world – a revolution. According to them, communities and electronic entities were created from the free flow of information and the creation of connections in the electronic spaces in Cyberspace. They believe that, as in Gibson’s book (Neuromancer), the online communities and entities in cyberspace are a world unto themselves, the uniqueness of which requires the creation of an independent set of rules that regulate conduct in it.
On the other hand, there are those who hold the evolution approach and claim that “there is nothing new under the sun […] new and innovative technologies can be included or swallowed up in the world of old and contemporary law. According to them, law uses definitions that have been refined over hundreds or sometimes thousands of years in which they have been used. They are enough to apply and take control of the new technological reality”. The commentators of the evolution claim against the commentators of the revolution that behind the developing technology are real players and tangible commercial activity with actual results in everyday reality. Hence, behavior resulting from artificial intelligence may have consequences in the real world, and the rules applicable in the real world must be applied to it.
The resulting conclusion is that the proposal of those legal commentators to create a new legal environment will inevitably create internal contradictions between the virtual world and the real world. Therefore, the practical and effective solution that we believe should be adopted is to adapt existing legal principles to the new reality and propose amendments to the existing principles where this is necessary. That is, the preference of the evolutionary approach over the revolutionary one [2].
The legal risk of the owner and creator of an AI feature
Hence, in accordance with the evolutionary approach we presented above, when we examine the desired responsibility of the owners and creators of an artificial intelligence feature in relation to copyright violations that occur due to that artificial intelligence, we believe that the issue should be examined according to the doctrine of contributory copyright infringement that was detailed in case AA 5977/07 the Hebrew University v. Shoken (Nevo, June 20, 2011).
Contributory Infringement of copyright refers – as a rule – to a situation in which even a person who does not directly infringe copyright is found responsible for it, and this is based on his behavior and involvement in the circumstances of the case. These are usually circumstances in which one party does not commit the violation himself, but he enables and permits the infringing activity of another party, or even assists and encourages the same activity. He serves as a kind of “intermediate factor”, in the chain of infringement – in the range between the direct infringer (“the edge” infringer), and the right holder.
The recognition of contributory infringement does not create a new category of violation of the rights of creators. It only allows retribution from those who took part in doing the harm. As will be clarified later, recognition of this doctrine dictates the cautious and limited application of this doctrine, only when specific conditions are met and in limited to exceptional cases only. The ambition is to apply this doctrine, while minimizing the harm to users and the public space as much as possible, and hence the contributory violation must be recognized only in the limited cases suitable for this. For this purpose, the existence of three conditions must be examined, which can only be recognized as a contributory infringement of copyright if they are met in aggregate.
A. The one condition is the actual existence of a direct violation;
B. The second condition, the actual knowledge of the infringer-contributor about the direct violation committed;
C. The third condition is the existence of a significant, substantial and actual contribution to the commission of the violation.
Therefore, in light of these tests, it must be assumed that the legal risk of the owners and creators of an artificial intelligence feature in relation to copyright infringement by the artificial intelligence is low. This is similar to the situation where Microsoft has no responsibility for a situation where a person violates copyright using the WORD software.
This complex issue raises many additional questions, for example artificial intelligence software is not yet completely autonomous, and in order for it to function, it must be trained using data related to the desired output. If the data they use are works protected by copyright, and the use of this data is made without the consent of the copyright owner, could the person who “trained” the AI software be liable for copyright infringement? Is the entity that owns artificial intelligence responsible for creating computerized filtering in relation to the products of artificial intelligence? Is it even possible to perform such a filtration?
The legal risk of the parties requesting output from artificial intelligence
What is the desired responsibility of a third party who requests output from the artificial intelligence and uses this output in a way that infringes copyright? For example: on the artificial intelligence website www.craiyon.com if we ask craiyon to draw batman for us we will get the following result:
What is the legal responsibility of a third party who makes prohibited commercial use of this output?
It must be assumed that in this situation the legal responsibility of a third party who uses a copyright infringing product created by artificial intelligence is the same legal responsibility of a person who finds an image on the net and uses it while closing his eyes to the legal risks arising from this action.
To the extent that there is an action in order to comply with the conditions of Section 47 of the Copyright Law which states that “the person who performs an action in a work from the actions specified in Section 11, or authorizes another to do such an action, without the permission of the copyright owner, violates the copyright, unless the action is permitted according to the provisions of chapter 4”, this would amount to a copyright infringement taking into account the protections provided in the copyright law.
References:
[1] The scholar Naqvi examines in his article how current copyright laws in the United States may relate to a work created automatically by artificial intelligence software:
Naqvi, “Artificial Intelligence, Copyright, and Copyright Infringement”, 24 Marq. Intellectual. Property L. Rev. 1 (2020), p. 16-20.
[2] Drori, Werzansky-Orland, Yarkoni, Bar Sade, “Internet Laws in Commercial Law”, pp. 58-57 (2015)