One of the popular technologies in recent years is neural networks and the works made of AI neural networks with the main participation from AI. These works are created with a great deal of composition from AI, similar to Chat GPT’s ability to collect and paraphrase in creating literary products. So, for works created by AI neural networks, whose copyright will it belong to?
With the development of this new generation of modern science and technology, it is clear that they can solve complex problems, including creative tasks that were previously only possible for a human, natural person.
The genius of copyright perspective on work created by an AI neural network
In 1824, Ludwig van Beethoven began writing his 10th symphony, but he died three years later, leaving the work unfinished and becoming the regret of countless scholars and musical geniuses who wanted to continue his work, creating a complete work over the next two centuries.
197 years later, at the end of 2021, the efforts of countless great people throughout the history of world music have been successfully accomplished by an AI neural network, completing Beethoven’s unfinished work.
In addition to writing symphonies, existing neural networks can also generate other subjects. Prominent in the graphics industry is the Midjourney neural network system with the ability to create sharp, top-notch images with vivid graphics, just by simple user commands.
Some other well-known AI neural networks include:
- GPT (Generative Pre-trained Transformer): GPT is a Transformer-based deep learning model that is pre-trained on large amounts of data from the Internet. GPT has been used in many applications such as automatic text generation, image caption generation, and even music content creation.
- AlexNet: AlexNet is a famous convolutional neural network that achieved the first success in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2012. AlexNet paved the way for the development of later CNN models and has been widely used in image processing.
- VGG (Visual Geometry Group): VGG is a deep and simple convolutional neural network. VGG performed well in the 2014 ILSVRC and has become one of the important models in image processing. VGG has a uniform architecture and can be applied in many tasks such as image classification and image segmentation.
- ResNet (Residual Neural Network): ResNet is a deep convolutional neural network architecture with the important contribution of remaining blocks. ResNet solved the vanishing gradient problem and helped train deeper networks. ResNet achieved impressive results in ILSVRC in 2015 and has become one of the important architectures in image processing.
- LSTM (Long Short-Term Memory): LSTM is a type of regenerative neural network (RNN) that is capable of processing long-term information depending on the previous context. LSTM has helped to solve the information loss problem in RNN and has been used in many applications such as machine translation, speech recognition, and automatic text generation.
Copyright to work created by neural networks AI
Currently, the laws of most countries do not have specific regulations on products created by AI subjects, especially in the field of legal due diligence.
In the future, with the appearance and popularity of neural networks such as Midjourney and Chat GPT, the completion of specific legal regulations is inevitable. However, this process will take time and the internationalization of regulations will take even more time than the process of internalizing laws within a country.
Provisions on legal protection of intellectual property subjects, including copyright subjects, is of a territorial nature. This means that they arise in relation to the laws of a particular country when that country is not a party to or has made any commitments to international Intellectual Property Agreements.
In most existing laws, the criterion of whether a subject is legally protected as a work depends on the creative nature of the author’s work.
In Russia, creative products are created by an individual or natural person with clear proof, that individual or natural person will be certified as an author who owns copyright for intellectual creation activities.
Therefore, the main criterion for recognizing the result of intellectual activity as the subject of copyright is the creative nature of the work of its creator. Only one individual can be the author of a work. Another criterion is the presentation of a work in an objective form.
Thereby, who will be the copyright owner for the product created by the Midjourney neural network and other AI neural network software and applications?
Regarding this problem, many international experts have presented a remarkable point of view when determining the creative nature of works, which can be divided into two approaches: subjective and objective.
The subjective approach lies in the fact that the creative contribution of the author in creating the work reflects the personality of its creator, their exclusive style thereby creating an inimitable uniqueness.
The subjective approach does not focus attention on the expression of the creator’s personality in the work, but on its novelty, its originality, its appearance on the market.
In Russia, the Intellectual Property Law on copyright tends to lean towards a more subjective approach, i.e. creativity, which is characterized by the creator’s personality, has a higher point than other factors.
If we look at it this way and assume that the owner of the product created by the neural network AI belongs to the person/company that wrote the AI neural network software or the person who entered the commands to create that product, then:
- The creator of the neural network will not be the product owner because their contribution lies in creating algorithms/code on that basis for the AI neural network to operate and draw on its own, collect information according to user’s request. In many cases like Chat GPT, the software is only used to synthesize and paraphrase, so the creator of the software cannot be the owner of the product because they have created the software but do not have any personality present in the product.
- Users of neural nets, who enter commands for the software to create the product are also not the creators of the work, the owners of the copyrights to the products created because they do not have any contributions to creativity for the product created. The pictures and paragraphs created do not show any personality of the person who entered the command to create it.
Thereby, the AI neural network in many countries still does not have any clear regulations on the arbitration of copyright and authorship for the products it creates.
Considering the decisions of many law enforcement agencies in developed countries such as the United States, the United Kingdom, and Germany, the courts in these countries have also stated that there is no specific legislation on the matter.
In fact, in many countries, the law has specific provisions, but in the opposite direction, for example in Australia and many other countries, it is absolutely not recognized for the inventors to be authors of software such as AI neural network,… typical in the AI DABUS recognition dispute in many Australian courts by Dr. Thaler (In a new update in April 2023, the United States has continued to become one of the countries that do not recognize AI as inventors).
However, within the scope of copyright, if an individual uses a neural network selectively to support that individual’s ability to synthesize and create, the work completely created by neural nets can still be recognized as the work of that author.
For example, Ms. Kris Kashtanova – a new generation artist who makes full use of AI neural network software, has registered the copyright of the comic book created from images in Midjourney with the USPTO because she has created a story frame with content, coherent, logical development from those images, “powered by AI but not created entirely by AI”.
Initially, the USPTO granted Kris a registration, certifying that it was the first time a product entirely made by AI was registered for copyright in the United States. However, the USPTO later reversed its decision on the grounds that it did not receive information that this product was made from Midjourney. Kris then announced that Midjourney had made an offer to assist her in appealing the USPTO’s decision.
However, this incident has created an alarming problem, not only in the USPTO’s method of handling and granting copyright registration but also the fact that if an author like Kris does not actively disclose information that their work is created by AI neural network, will the USPTO not recognize and continue to grant registration as usual?
You can see a list of International IP Firms here.