Technology has become an indispensable part of everyday life and vital to society because of the risks and opportunities it presents, and as a result, it is often subject to regulation. Adopting new technologies in society can have various consequences, thus requiring the development and implementation of special laws.
Such philosophical questions as whether artificial intelligence is able to go beyond and become a full-fledged consciousness, making independent decisions, have been raised more than once. Even in movies more increasingly have begun to grant the concept of artificial intelligence and assign it a force beyond the control of mankind. For example, director Roland Emmerich’s recent film “Moonfall” shows the dark side of artificial intelligence, and its unbridled power. The uncontrollability of AI, created and brought to its “peak” by humankind, soon leads to terrible consequences, which threaten the end of the world for the earthlings. But there is undoubtedly a positive reflection of the acquisition of intelligence in AI as ideas for movies. For example, Steven Spielberg’s famous vision, 2001’s Artificial Intelligence. Here, by contrast, robots are shown in a better light and are even capable of showing feelings, such as unselfish love and sincere empathy. One can be sure that this theme will never cease to be relevant in this field.
Since the steam engine, there has been a long tradition of regulating technology in society through legal mechanisms. Thus, legislation has a significant impact on innovation and how well technology develops in a particular area. It’s hardly a surprise to you that regulatory problems have arisen in many jurisdictions since the first AI-based services became available. Despite this, regulators have not kept up with innovation, and based on “yesterday’s” laws are establishing regulations for floods, although these regulations are somewhat outdated and need to be improved, as well as the development of a new set of laws regulating this area.
To talk about the regulation of artificial intelligence in copyright law, it is necessary to have a general notion of what these terms mean and the problem that may arise.
Artificial intelligence is defined as the modeling of human intelligence processes by machines, especially computer systems. Artificial intelligence can be applied in a variety of sectors, can improve existing technology, and can generate novelties. Accordingly, AI can be used for numerous actions in society, in a number of different tasks, including detection, data management, communication, action, and recommendation. Thus, AI has become a major lever for the functioning of several applications, such as chatbots, intelligent assistants, facial recognition devices, and automated driving. Automated systems can influence or even determine important aspects of people’s lives, including health care, employment, housing, and education. Here, for example, the recently created ChatGPT neural network is already integrating into sectors ranging from education to generating the perfect Tinder rolls. It’s funny, but this kind of activity in Tinder doesn’t put anyone off yet, but at this rate, we may quickly get to a point where ChatGPT will communicate with itself in Tinder.
In other words, a computer program designed for machine learning purposes has a built-in algorithm that allows it to learn from input data and to evolve and make future decisions, which can be either directed or independent. When applied to works of art, music, and literature, machine learning algorithms actually draw lessons from the input data provided by programmers. They learn from this data to create new works, making independent decisions throughout the process. An important feature of this type of artificial intelligence is that while programmers can set parameters, the work is actually generated by the computer program itself, called a neural network, in a process similar to human thought processes. But, every coin has a flip side: neural networks in this context are not equally exempt, because of the potential threats and risks they carry with them. So what threats and problems do they bring with them? – is a pressing question for the near future, which we will explore with you.
- Unpredictable behavior:
Artificial intelligence systems are designed to learn from the environment and make decisions based on that learning. However, this can lead to unpredictable behavior because the system cannot predict exactly how it will respond to certain situations or data. This can lead to unexpected results and potential hazards.
- Job loss:
AI is increasingly being used in various industries, including manufacturing and healthcare, to automate processes and replace human labor. This can lead to widespread job loss, especially in areas with a high concentration of unskilled labor.
- Security Risks:
AI systems are vulnerable to hacking and cyberattacks, which can lead to the theft of sensitive data or manipulation of the system for malicious purposes.
- Biased results:
Artificial intelligence systems are only as good as the data on which they are trained. If this data is biased or incomplete, it can lead to incorrect results that reflect this biased data.
Can you use copyrighted data to train artificial intelligence models?
The use of third-party data that is necessary for the deep learning process is the first problem. According to copyright law, all such data must first obtain the consent of the owners of the relevant rights. However, getting permission for all of the data that deep learning services use is sometimes nearly impossible.
Although it is not regulated, there is Stable Diffusion software from Stable AI, which continues to copy billions of copyrighted images to allow AIs like Midjourney and DeviantArt to create images in the style of these artists without permission, and this becomes the cause of lawsuits filed by artists. Speaking of which, artists from various states recently filed a lawsuit against Stability AI, Midjourney, and Deviantart for a blatant and enormous violation of their rights.
How do governments deal with copyright ownership of computer-generated works?
The copyright of works created by artificial intelligence is the subject of the second problem. Who owns the copyright in contracts or other legal documents written by artificial intelligence if they have commercial value? In such circumstances, a dispute may arise between the supplier of artificial intelligence services and the consumer of that service over the ownership of any material created. Let’s take a look at the regulation of this phenomenon in some countries.
Looking at Canadian practice, not long ago technological neutrality in Canada was affirmed by the Supreme Court of Canada as «recognizing that, absent Parliamentary intent to the contrary, the Copyright Act should not be interpreted or applied to favor or discriminate against any particular form of technology». Thus, it is necessary to apply the skills and judgment test to works created by artificial intelligence, consistent with other ways of creating copyrighted works. If an artificial intelligence automatically uses a vague and highly randomized data selection process, it becomes difficult to argue that its user has exercised the necessary skills and judgment to be considered the author of the final product. Following the Supreme Court of Canada’s reasoning in CCH Canadian, a user who simply presses a button to engage the algorithmic processes of artificial intelligence would likely be engaged in a “purely mechanical exercise.” Consequently, such a user should not be recognized as the author of the work. Such a course of reasoning could prevent the creator of artificial intelligence from being the author of works derived from highly automated processes.
In addition, technological neutrality may play a role in recognizing situations where there is joint authorship between the creator of the AI and the user of the AI. There is no doubt that the creator of the AI would be the author of the AI code. This should not automatically extend to authorship in AI works, especially if the creator of the AI makes no contribution to the data selection process. However, such a clear-cut scenario is almost unlikely. The creator of the AI is likely to collaborate with the AI user in the development of the AI that will generate the work.
In these cases, there may be a need for joint authorship between the two, since they have both applied their skills and common sense to create original work. But this approach is problematic because it would present a dilemma in which large teams of people are involved in the design and use of artificial intelligence. Thus, in the absence of legislative reform of the Canadian Copyright Act, the law is likely to remain confused about the author(s) of works created by artificial intelligence. This dilemma underscores the commercial importance of carefully crafted written agreements to identify the owner of works created by artificial intelligence.
United States of America
U.S. legislators have been talking for years about the prospects and dangers of AI. But, as with previous waves of technological innovation, the speed with which products come to market has far exceeded the government’s willingness to regulate.Regarding the first issue in the U.S., the use of training data amounts to illegal reproduction, which means an infringer can be held liable for copyright infringement. District courts are divided on this issue. However, even if there is infringement during machine learning, the use of copyrighted materials to train AI is likely to be allowed under the “fair use” doctrine.
What is the “fair use” doctrine?
This is a legal doctrine in the United States that allows limited portions of copyrighted material to be used under certain circumstances without obtaining permission from the copyright owner or paying any licensing fees. Fair use applies to all kinds of works, including text, images, videos, and music. Some examples of fair use include teaching, scholarship, research, criticism, commentary, news reporting, and parody.
The most common question about fair use is how to determine how much of a work can be legally used. Unfortunately, the law and subsequent court cases provide no clear guidance as to which specific uses are fair and which are infringing. There is no formula, percentage, or amount of work that automatically counts as fair use.
“Hey, what’s going on on the legislative side in the U.S.?” you ask.
Here’s a little digest from Congress, where lawmakers have proposed rules for facial recognition and other AI applications.
- The White House has an AI research office and research office has released a draft AI Bill of Rights.
- The Federal Trade Commission, the Equal Employment Opportunity Commission, and other federal agencies have begun introducing new rules for the use of AI.
The Draft Artificial Intelligence Bill of Rights includes these five points:
- Automated systems must be safe and efficient.
- Users should not experience algorithmic discrimination.
- Users must be protected from data misuse and have control over how their data is used.
- Users should know that an automated system is being used and understand how it affects them.
- Users should be able to opt out and, if possible, talk to the person. Users should be able to contact a real person if the automated system malfunctions, gives an error, or they want to challenge a decision.
Currently, as far as permissive use is concerned, there are no rules in the South Korean Copyright Act regarding AI training data. Consequently, when copyrighted materials are used as AI training data under current law, such use could presumably be considered copyright infringement. However, Article 35-3(1) of South Korean law states that “…if a person does not unreasonably infringe upon the legitimate interests of the author without coming into conflict with the ordinary use of works, he/she may use such works. In other words, the doctrine of fair use is also possible under South Korean law, and there is a possibility that this provision will be applied to artificial intelligence training.
On the issue of ownership, South Korea’s Copyright Act defines “work” as “a creative work expressing human thoughts and emotions” in Article 2(1), and “author” means “the person who creates the ‘work'” in Article 2(2). Thus, unless the current law is amended, only human beings can be considered authors of creative works. Thus, works created by means of artificial intelligence are not protected by existing law
However, South Korea could also see progress on this front. The Presidential Intellectual Property Council formed an Ad Hoc Expert Committee on AI Intellectual Property in June 2020 to develop a nationwide AI policy. This committee will discuss many legal issues, such as:
1) whether AI should be recognized as an author;
2) whether works created by AI should be protected at the same level as human works;
3) who owns the work created by AI.
Terms of Service
But, let’s move away from the topic of AI regulation a bit and look at the user side of neural networks. Perhaps you’ve been thinking about becoming a paid subscriber to Midjourney, but would like to better understand what you’re allowed to do and what you’re not or do you want to be on the same page in a fast-paced age, in this case guide below will definitely be helpful to you!
Midjourney is an art development neural network whose goal is to create a collaborative space where everyone creates different works of art to share with the community.
In other words, Midjourney facilitates the sharing and recycling of other users’ Midjourney images, as it helps to constantly improve artificial intelligence over time.
Let’s take a look at the public offer of Midjourney
Subscriptions in this neural network, as in many things, are divided into 2 types: Paid and Free.
Non-paying users are granted a Creative Commons Noncommercial 4.0 asset license, which means they can use images for personal projects as long as they specify visible attribution.
Paying users are classified as “owners” of the assets they create and can use them for any commercial purposes they choose.Despite “owning” the images, Midjourney’s terms also implicitly “compel” you to allow someone else to use your publicly available images.
- Perpetual, non-exclusive, and worldwide rights.
- Rights to sublicense, reproduce, and prepare derivative works.
- Rights to publicly display, perform and distribute images and text cues.
- You “own” all assets created with Midjourney.
- However, they may still be used and remixed by others.
- You may use the images for commercial purposes as long as you are a paying user.
We know some of this may seem confusing, but essentially it says that anything you create with Midjourney can be used by Midjourney and anyone else. And depending on whether you are a free or paid user, you are given different levels of usage rights.
As we can see neural networks bring with them not only opportunities but also potential threats, not only lawmakers and authorities intend to regulate AI. Developers are concerned about this too, but at this stage, most AI developers adhere to the fair use doctrine, but it is important to consider the fact that in the United States, the Federal Trade Commission is already in the process of creating new rules of commercial surveillance and data security, which will regulate any company that develops and deploys artificial intelligence systems. Individual states such as Massachusetts are also mulling the bill. And South Korea’s AI Expert Committee began drafting a neural network policy even before the push from new AI (ChatGPT, Midjourney).
And finally, we’d like to share with you ChatGPT’s response to the question we asked about the legal practices that will be in demand in the near future. Below is the response from the AI :
“It is anticipated that the legal practices that will be most in demand in the near future are those that are related to the use of artificial intelligence (AI) and machine learning. This includes the development of legal frameworks to regulate and control the use of AI, as well as the creation of legal solutions to address the ethical implications of AI. Additionally, as the use of AI for commercial purposes becomes more widespread, there will be a need for legal practitioners to advise on the use of AI and to provide legal guidance on issues such as data privacy, intellectual property, and liability.”