Copyright and Neural Networks: Current Issues

The topic of neural networks, the opportunities and risks of their use, whether GPT chat can write a symphony or DALL-E take a blank canvas and turn it into a masterpiece is all over the news these days. Neural networks have "seized" podcasts, startup pitching sessions, blogs and webinars - it is being actively discussed whether they can be used to reduce costs and write source code, create artwork.

Meanwhile, the legal status of the results issued by the artificial intelligence remains uncertain: on the one hand, a work is created, but on the other hand - the question arises as to whether this result has a creative element and emanation of the author's personality, which is the key characteristic of copyrighted works, if so, who owns the rights to this work there is no clear answer, but let's try to go deeper.⁠⁠

What are neural networks?

At the moment, neural networks have gained immense popularity among users because of their capabilities and flexibility with respect to solving specific tasks.

A neural network (or artificial neural network) is essentially a computer system that imitates the nervous system of living beings to process information. Neural networks are used in computer-assisted learning and artificial intelligence for data processing, pattern recognition, classification, forecasting, and many other tasks.

Neural networks can be very effective in a variety of tasks, including computer vision, speech recognition, natural language analysis, and much more.

Who owns the right to the result created by the neural network?

The authorship of content created by neural networks, as well as the possibility of protecting it, is currently not regulated at the legislative level and is specified by the developers in the Terms of use.

From the legal perspective, AI cannot be the author and have corresponding rights. Consequently, the subject of copyright may be either the developer of the AI, or the user providing the input data.

There is no unified approach to the recognition of authorship. However, most jurisdictions adhere to the substantial contribution approach, in other words, the recognition of authorship for a subject is possible if substantial creative effort has been made to create the work. For example, in the United States, the U.S. Copyright Office, has recognized the possibility of registering a work containing human-created elements combined with the use of artificial intelligence, but on the condition that the AI was only a tool for creation, and the substantial contribution remains with the human.

In general, the authorship of works created with the help of artificial intelligence is quite a complex issue, and requires the development of new legal rules and principles to take into account all aspects of the creation of such works.

In order to eliminate potential copyright disputes, the developers enshrine the terms of cooperation with users in the Terms of Use Agreement.

As an example, take the most famous resource ChatGPT, in paragraph 3 (a) of the Terms of Use, it is defined that the copyright on the input data belongs to the user in accordance with the law. In turn, OpenAI assigns the copyright on the response created by the chatbot to the user.

Therefore, when using ChatGPT, the copyright for the created content will belong to the user, however, we emphasize that the Terms of Use Agreement does not change the law. Therefore, if in a particular jurisdiction, for example, the U.S., there are requirements for de minimis contribution of AI to the final result and for mandatory substantial human contribution, then such provisions must be taken into account. 

How can neural networks be used?

Content created by neural networks can be used for any purpose, including commercial ones. Neural networks are most often used to create art, icons, texts for a website or individual articles, and source code. All of the above are subject to copyright and carry certain risks of using artificial intelligence to create such objects.

The person recognized as the author, i.e. the developer or the user, depending on the specific situation, is responsible for the violation of the exclusive rights that may potentially be violated by the neural network. In the course of information processing the neural network may use the copyrighted works, or when generating a response it is possible to reproduce the protected object, which may be considered an infringement of the exclusive right. 

Therefore, companies must take measures to reduce the risks when using neural networks. For example, for companies hiring third-party developers to write code, under current circumstances it is necessary to include a new clause in the contract for software development, distributing the risks of using neural networks and possible violations associated with its use. Such a clause in the contract can prevent possible disputes arising from such relationships.

In addition, for legitimate use with respect to the original data and works that are processed by a neural network, the rights of the author of the original work against processing and copying must be respected. This issue is most acute for users of neural networks interested in code generation. If a company uses AI to generate code, it is advisable to run the resulting code with open source licensing tools to determine the origin of the code and verify its license. In turn, the code written solely by a neural network without human involvement, depending on the jurisdiction, may not be protected by copyright, for the reason described above: due to the absence of a "human" creative element. 

Consequently, when publishing, distributing, or otherwise using objects created by a neural network for commercial purposes, it is necessary to check in advance whether the exclusive rights of other authors are not violated.


 One of the presented paragraphs was written using the ChatGPT neural network.