AI technology has been around for a while, but it has only recently started to become as accessible a tool as Google. ChatGPT, DALL-E 2, Bard and many other chatbots will soon be available to everyone, offering the prospect that the technology will replace many professionals in a heartbeat.
Every student will be able to ask his preferred AI software to write his university assignments; every journalist will have his column ready for deadline without effort or research; and a lawyer’s work in many areas will be replaced by AI technology.
But where does all the information come from? Do we really believe that computers can replace law school graduates? Can AI design our homes or create original artwork? These are questions that also raise the potential for a conflict between technology and copyright.
When we ask our computer to draw a picture or draft a lease agreement, the computer obviously needs to search databases that are available to it in order to generate any new product. This introduces two different questions: who owns the copyright in what the computer creates? Does the new creation infringe on someone’s copyright? The answer could be argued in several ways.
It is clear that for the machine to write or draw anything, it needs access to existing information, documents, research, drawings, pictures, etc. In simple terms, it needs examples to copy from or – in the case of artwork – to be “inspired” by. This access allows the software to generate works by using its computing abilities to retrieve pieces of data from already available databases in order to create a new creation/work.
Who owns the newly generated work?
Copyright law varies by jurisdiction, but generally, the person who creates an original work is considered the copyright owner. In the case of AI-generated works, it can be difficult to determine who the creators are. Some countries, such as the US and Israel, have laws that provide that computer programs, including AI-generated works, are considered “literary works” and are protected under copyright law.
In these cases, the person who writes the code for the AI system is typically considered the owner of the copyright, unless there is an agreement to the contrary, as for example in an employment agreement. But is the copyright owner also the owner of the copyright in the resulting, newly generated works?
This could be a “simple” answer, as in the case of the person who uses the AI software to generate the new work being the copyright owner. But it’s more complicated than that, because the software uses multiple sources to create the end result. In such cases, the copyright ownership may be split among multiple parties, or may be determined through a contractual agreement.
In copyright terms, this is a derivative work, which means that the original author, the creator of any of the work used by the AI generator, has some copyright in the final work generated by the machine. There have not been many court cases specifically addressing the issue of copyright ownership of AI-generated works.
THIS IS new, uncharted territory. However, some courts have applied existing copyright principles to analogous situations involving computer-generated works, and these cases may be instructive.
For example, in the case of Feist Publications, Inc. v. Rural Telephone Service Co., a European court held that while a computer program may be protected under copyright law, the resulting output is only protected if it is original and creative, meaning that it owes its origin to human authorship. This was 10 years ago.
Such cases suggest that even in the absence of specific court decisions addressing AI-generated works, existing copyright principles may still be applied to determine the ownership of such works. Hence, the person operating the computer that looks for data and asks the questions leading to the creation of the work should be the owner of the work, pending the rules concerning derivative works that provide copyright protection to the original creator, as well as rights to the creator of the derivative work.
Another issue involving derivative works, and thereby leading to the question of infringement, is that when using the original work to create a derivative work, the permission of the original creator is required if the original work is still protected by copyright and attribution to the original creator is also required. This certainly makes the issue more complicated.
If it involves “permissions” and “multiple copyright owners” – will we need to give credit to all sources of information? Is it any different from academic research where the student uses several articles to create his own work? Can the computer follow the citation rules? In the next few years case law in different jurisdictions and with different facts will shape the law on the subject and create new “rules.”
Do AI works infringe copyright?
AI-generated works may infringe the copyright of others if they use protected elements from existing works without permission. Copyright law provides exclusive rights to creators, including the right to reproduce, distribute and display their works. When an AI system uses existing works without permission, it may be considered as copyright infringement.
A more precise answer could be considered in respect to a specific machine-created work, by taking into account the number of original works used by the machine in order to create the newly generated work and the need to give the proper attribution/credit to the original creators.
IT IS IMPORTANT for AI developers and users to be mindful of copyright laws and to ensure that their AI systems do not infringe on the rights of others. This may involve obtaining permission from copyright owners, or seeking legal advice to determine whether a particular use of a copyrighted work is permitted under “fair use” or “fair dealing.”
We are at the beginning of an era involving technology and law. We are starting out from the belief that the simple use of copyright law will suffice, but as technology evolves and improves, it will be harder to identify the sources of the data used by the machine, and case law will become more complicated.
Alternatively, new laws may require that the AI software disclose all its sources and/or give automatic credit – citations.
The user of the AI machines will be the owner of the copyright in the newly generated work, but will he have partners in the work – partners who have not given their consent and are “forced” by the machine to contribute their works or parts of them?
The writer is a partner at Firon law firm, heading the intellectual property department.