*Originally published in Estadão.

**This is an AI-powered machine translation of the original text in Portuguese.

Ever since the first websites and applications using Artificial Intelligence (AI) emerged, countless questions have arisen regarding the subject. Last week, Senator Rodrigo Cunha (Podemos/AL) introduced Bill No. 3592/23, which establishes guidelines for the use, through artificial intelligence (AI), of images and audio recordings of deceased individuals, with the aim of preserving their dignity, privacy, and rights even after death.

The subject is complex and subjective. While some people may view the use of AI as a way of paying tribute to someone’s image and work, others may consider it inappropriate and a distortion of the deceased person’s “persona.”

The desire to connect with deceased individuals is nothing new. This theme has been explored through different art forms over time, ranging from fantasy stories to technological dystopias. Recently, the debate resurfaced with the digital recreation of singers such as John Lennon, who died in 1980, and Elis Regina, who died in 1982. However, around a decade ago, the series Black Mirror (2013) had already addressed the reproduction of posthumous personas in one of its episodes. In it, the protagonist uses an online service to “communicate” with her deceased husband through a digital reproduction created from his images and social media posts. What stands out in these recent cases is that the popularization of Generative AI appears to be bringing us closer to something that had previously been confined to fiction.

It was partly for this very reason that a bill addressing the use of AI to generate content featuring deceased individuals—Bill No. 3592/23—was introduced on the 19th.

The use of generative AI requires caution, including reviewing the work produced to ensure its accuracy, verifying sources, and respecting third-party intellectual property, among other considerations. When it comes to creating content that simulates what might have been produced by someone who has already died, even greater care is required, as this raises issues that extend beyond the technical aspects of AI and even encompass philosophical concepts related to life and death. Therefore, the subject is naturally complex and subjective: while some people may view the use of AI as a way of paying tribute to someone’s image and work, others may consider it inappropriate and a distortion of the deceased person’s “persona.”

The text of Bill No. 3592/23 makes it clear that this type of content may be prohibited either by the individuals themselves before their death or by their heirs.

Before using any form of Generative AI—for example, to create a digital image or replicate a person’s voice—it is important to obtain the appropriate authorizations. This applies to everyone, whether deceased or living. Indeed, obtaining authorization through voice and image release forms was already standard practice in audiovisual sectors long before discussions about artificial intelligence became widespread. Therefore, in the case of deceased individuals, authorization from their heirs is required for their image to be reproduced. This was also the position adopted by Brazil’s Superior Court of Justice (STJ) in a case involving the use of the images of Lampião and Maria Bonita in an advertising campaign—for commercial purposes—without authorization from the couple’s daughter.

Under Bill No. 3592/23, the proposal is that heirs should be able to decide whether Generative AI may be used to represent a deceased person, including the right to refuse to allow the person to be represented by such means.

And what about the artists themselves whose likenesses are being—or will be—reproduced by AI? This is a delicate issue because, although the reproduction of photographs and videos of deceased individuals is a longstanding matter, the use of Generative AI to create new contexts makes the problem more complex. The Civil Law Conferences (Jornadas de Direito Civil) have already addressed image rights after death, but the practical application of their statements to Generative AI still needs to be discussed. Factors such as the public prominence of the person portrayed, the truthfulness of the facts, and the nature of the use—commercial, informational, or biographical—must be considered, along with, of course, the wishes of the person portrayed by the AI.

There are notable examples of public figures who have spoken publicly about this subject. In late 2022, it was reported that actor Bruce Willis had sold his image rights to allow audiovisual content featuring him to be produced using artificial intelligence, thereby expressing his wishes regarding the use of his image in this manner. Conversely, actor Robin Williams prohibited the use of his image for 25 years after his death. Although he died in 2014, this decision may have been motivated by the computer-generated imagery (CGI) effects already available at the time.

The text of Bill No. 3592/23 places the individual’s wishes at the center of the issue. Therefore, if a person has refused to allow the use of Generative AI, that decision must be respected. The role of the heirs would then focus on two other situations: when the person authorized such use, or when the person did not explicitly express a preference either way. In cases of explicit authorization, the heirs would be responsible for overseeing the use of the deceased person’s image and could, for example, reject applications of Generative AI that they consider inappropriate. If the deceased person left no statement on the matter, the decision would rest with the heirs, although this could be a complicated issue in relation to people who died much longer ago.

Despite these provisions in the bill, the further back we look, the less likely we are to find any indication of what these deceased individuals thought, since the possibility of using Generative AI was not even imagined in their time. In such cases, the creation of scenarios through Generative AI becomes more context-dependent. For individuals who died before the development of the technology that enables such reproductions, what could be considered an “intention not to allow the use of their image after death,” as stated in Bill No. 3592/23? AI is capable of doing more than merely reproducing content created by a person during their lifetime, and this nuance will need to be taken into consideration.

No single measure will, by itself, prevent risks in every case. Generative Artificial Intelligence technology has only recently become widespread and gives rise to a variety of opinions, while new applications for it are constantly being explored. Nevertheless, transparency regarding its use is a central element in the application of this technology.

It is important to know when particular content has been generated by AI and to distinguish it from other audiovisual works created by people. This helps prevent confusion among members of the public when they encounter the use of Generative AI. By understanding how Generative AI is used, we can better understand its capabilities and limitations. Reactions to Generative AI today are already quite different from what we saw in the well-known case of “Pope Francis in a white puffer jacket” just a few months ago. This is due, at least in part, to the public’s increased understanding of how the technology works. Bill No. 3592/23 takes the same approach by clearly providing that any advertising created using AI must prominently inform consumers with the wording “advertising using artificial intelligence.”

Transparency also plays an important role by enabling closer oversight by regulatory authorities. This makes it possible to understand how Generative AI is being applied, identify more and less successful cases, and guide stakeholders in the AI field toward the ethical, safe, and responsible use of this technology over time.