The debate surrounding the ethical and legal use of Artificial Intelligence (AI) gains prominence with the proposal for Bill 4025/23, presented by deputy Marx Beltrão. Amid discussions about copyright, image and intellectual property, the initiative aims to bring clarity to delicate issues involving the creation and use of works generated by AI. In this context, lawyer Thereza Castro , focused on copyright and innovation strategist for the literary market, brings valuable insights into the legal and ethical implications of this constantly evolving scenario.
Could you explain how Bill 4025/23 proposed by Congressman Marx Beltrão addresses the use of artificial intelligence in relation to the image rights of people represented by AI?
Bill 4025/23 proposes that the manipulation of a person’s image by artificial intelligence depends on their express authorization, that is, the person’s agreement must be expressed so that it is possible to manipulate their image. It should be noted that we are talking about the act of manipulating, firstly, we do not even get to the patrimonial consequences of the product. Just to create manipulation over someone’s image, whoever it may be, it is necessary for the person to authorize it. Furthermore, the project states that when it comes to a deceased or absent person, the list of those who could give this authorization would only be the spouses, ascendants and descendants.
How do you see the importance of prior authorization for the use of people’s images by AI, especially in the context of the emergence of deep fakes, as in the case of the Volkswagen advertisement with Elis Regina?
Artificial intelligence brings us new challenges, so it is necessary, thinking about everyone’s safety and the good use of the tool, that legislation follows and embraces the possibilities of innovation, hence authorization, which is just a small step in the middle of the sea of issues around AI, is essential.
We must remember that previously the authorization to use an image of a person, to be captured, or use an existing one, already existed, but AI has other horizons, making it possible to produce images of the person that never happened, as in the case of the advertisement with Elis Regina .
Although the use made by Volkswagen in the advertisement, as said at the time, had the authorization of the late singer’s family, it draws attention to the reliability that AI is capable of producing. And when we think about it, regulating authorization is just the tip of a huge iceberg, which we should have overcome a long time ago. There is an urgent need to regulate the use of AI widely, only in relation to the image we are talking about the concrete possibility of ethical and non-ethical uses of a tool that can make an image realistic enough to inspire belief, which is a potential danger in different levels.
remuneration for copyrighted works used to train generative AIs . How important is this proposal and how could it affect copyright holders?
This is also a very important point, since generative AI needs to use a database to generate its responses, and train itself each time for greater assertiveness. In other words, if you use a copyrighted book to ask the AI to generate a response, the conceptual explanation of which is written there, there is no remuneration involved for the owner.
It seems basic to say that if I want to use someone else’s work I need to pay for that use, if the owner has not made it publicly available, right? But this is not what is happening at the moment, so this point of the proposal is of utmost importance for the holders, who, after all, must be remunerated for what they produced as authors, including, with this rule in force in the future, the companies that own the AIs will need to clarify and classify the data they use, giving more clarity to what they are using, in fact, as AI once trained will not forget what it has learned, it is reasonable for us to think about the period in which such works are used without compensation.. Rules should come from this provision for it to be applied effectively.
Regarding the copyright of works generated by AIs , the bill establishes that they do not enjoy copyright protection. How do you see this decision in terms of protecting the rights of the creators of these works?
This point is actually, in my opinion, the fallacy of the project. Because it maintains the provision of the Copyright Law (9,610/98) by establishing that only a human person can be an author, since we are talking about the creation of a soul. He also maintains the point that the legal entity may enjoy the protection of the terms of the aforementioned Law, which occurs when a company or organization represents the collective of authors, for example. However, works produced through AI are being widely published by people who have given it a command. Would these works then be owned by these human beings? The project does not give us the answer. And if we think that AI actually uses a database where we have works actually produced individually or collectively by human people, without the use of AI, works protected by copyright, this becomes even more cloudy, as the generation of Wouldn’t the product perhaps use excerpts, albeit modified, from these works? And this, and many other answers depend on the interpretation of what AI is in the end and what influence the person issuing the command has, or does not have, over the generated product.
In short, this is not exactly what the project provision says, it just says that the generated product cannot be attributed to AI authorship, but it does not determine whether or not it could be attributed to the person issuing the command. So, I return to the question, who are you calling the creators of these works? Because this answer is also embedded in a thought about what AI actually is and the role of those who give it commands .
Starting from the point that the product generated by AI does not enjoy copyright protection, and that the human controller is just pressing keys, then everything generated by AI would be free to use, which seems like a problem for those who in some way moment may have understood differently when purchasing a license.
On the international scene, there are different approaches to dealing with products generated by generative AIs . Could you explain the main schools of thought and how they relate to the proposed bill?
Yes, there is no unanimity on how to deal with AI, just as there is no consensus on how to deal with the product of your generation. Europe, with its recent legal changes in what is being called the European AI Act, is the one that seems to be actually discussing this widely and seriously. We understand that today there are different views on AI and the role of the human being who uses it, when it comes to the rights over the product generated by this interaction.
One of the currents proposes that although the authorial creation is from the human soul, patrimonial protection would remain for the holder of the tool, so the contents should be licensed. Here we have an understanding that the developer created a complex machine capable of delivering creative results, and the human being’s commands do not have as much influence on this point, being just a press of buttons.
In the second view, we have a general lack of copyright protection, neither for the machine nor for the human being who gave the command. This initially seems like a trend followed by the project. In this case, the machine is devoid of a soul, however complex it may be, and the human being has no fundamental contribution of soul and creativity to the final product . We could then understand a free use of the result.
A third party also does not grant copyright protection, but limits the product to general use while understanding that the holder would be the human being or company that assigned the commands and paid for the software license, this would then be the one that could exploit the result generated by the AI. In this vision, we have AI as a tool, close to what we see for other software, such as an office suite, we pay to use it , but what we generate in it is, in the first instance, who is paying for its use.
Do you agree that analyzing the rights involved in AI products would be more effective if discussed on a case-by-case basis? How could this affect the safe use of AI in various areas?
I understand that there will be cases in which individual analysis will be necessary, the point is extremely complex and its consequences may be even more so, therefore, even legislation that tries to regulate all possibilities will fail, but it must exist to guide the analysis of case will undoubtedly arise as a necessity, when the guiding rules are not sufficient for the factual situation that arises.
What we cannot, however, is operate in the absence of these general rules, which will only come through a broad debate and would be about the technology that is already presented to us. Although the existing legislation is broad, it does not cover many of the possibilities and challenges that AI already gives us, which creates insecurity not only due to the lack of clarity in the application of what already exists about this new technology, but also due to possible uses of loopholes legal for what is currently not achieved by existing laws. Let’s remember one of the fundamental principles of our Magna Carta that aims to bring security, but that also obliges the legislator to be in accordance with the times that present themselves, art. 5, II of the Brazilian Federal Constitution states: “No one shall be obliged to do or refrain from doing anything except by virtue of the law”.
How important is it to establish clear rules for the use and exposure of AIs in relation to copyright and intellectual property rights?
It is important for us to have a safe, ethical and transparent use of this technology, ensuring that everyone knows what they are using, that the databases do not violate rights, how and when they can or cannot use the products generated by the AI tool, that if it is used, when presenting a product with its use, it is clear that this occurred, not confusing the creation with this interference from those who did not use it. In other words, the rules will bring security and fair treatment to everyone involved.
Do you believe that the evolution of AI technology makes it more difficult to analyze the rights related to the products generated by it? How does this affect the professionals involved and the rights to creation?
Without a doubt, as AI improves it becomes more difficult to detect without clear exposure that the tool was used whether in creating images or composing text, for example. If this is not clear, we give the possibility of evaluating two different things with the same weight. It is important, when presenting a work, for example, that the rules are the same, so if there is the possibility of using AI it must be extended to everyone, if not then no one can use it. This assumption ensures that everyone feels fair, treated the same way, and starting from a more or less equal point. If we do not establish previous standards, someone could present a product as their own and, having a highly developed AI, asserting or detecting the use of AI would be practically impossible without a very in-depth analysis, which would convey false information and give an unfair advantage.
The recent case of the disqualification of a work from the Jabuti award due to the use of AI to generate its cover raises questions about what we want to encompass in relation to the use of AI. How do you see this case in terms of the future of AI use?
The case is an illustration of what I mentioned in the previous question. If we are evaluating a cover that was theoretically generated only by human creativity, it is not fair that we have one that used the help of the machine evaluated with the same criteria. I understand that the disqualification decision was made in order to guarantee an equal and fair view of the competitors, and it already demonstrates the difficulty in detecting this use, allowing this to not be seen until an advanced point in the evaluation, this highlights the pressing need of rules, and also demonstrates that possibly, having these rules, we will have a distinction between what was done with the machine and what was not, guaranteeing equitable treatment in similar situations.
Considering the complexity and potential ramifications of using AI, what steps do you suggest to improve the legal and ethical treatment of AI in the coming years?
It is essential that we discuss as a society our understanding of AI and discipline its use, this will ensure that we are able to deal with the future developments of technology, which are already presenting themselves very quickly, with a minimum of direction, so that from time to time It is time for us to look into this topic again, and remove, as much as possible, the uses that violate the individual and fundamental rights of human beings.
We must discuss and regulate this point so that we can innovate with its use in a safe way, knowing what will become of the new creations, what can be done or what points we should pay attention to, and so that we can quickly prevent mistaken uses of technology.
Furthermore, as long as we remain at a standstill, the ethical discussion also remains at a standstill, leaving our current society and the future generations that are being educated today, unprepared to face this future, or living with a future without clear guidelines, a future that is rushing so quickly towards us. .
Follow Thereza Castro on Instagram