Although Artificial Intelligences (AIs) are commonly associated with the chat models currently available in various corners of the internet, they are also present in our daily lives in many other ways. Often poorly understood by the general public, the need to comprehend these tools beyond their most common use has opened the door for an interview withPopular PerspectivesThis week.
In addition to understanding how AI works and how it relates to the geopolitical rivalry between China and the United States, the conversation covered the proposed regulation of this technology in Brazil and its impact on democracy and the 2026 election process.
To discuss this topic, we invited Ettore Medeiros, a doctor and professor of Social Communication at the Federal University of Minas Gerais (UFMG), who has been researching the field of technology and artificial intelligence.
"There is an infinite number of possibilities to re-imagine what artificial intelligence is. Currently, our imagination and understanding of AI is very limited to what foreign large tech companies are offering us. And that's a big problem because we are losing our chance to imagine other futures with our AI, based on our needs," the professor points out.
Throughout his academic journey, he has also researched, among other topics, gender and sexuality studies, digital platforms, branding, digital marketing, diversity, and inclusion.
Check out the full interview
Brazil of Fact MG - What are Artificial Intelligences and where do they fit into our daily lives?
Ettore Medeiros –Artificial intelligence is an area of computing, with the goal of making machines develop skills that resemble what we, humans, flesh-and-blood beings, are capable of. Whether it's recognizing an image or establishing something that simulates a conversation, even though chat AI's are not, in fact, conversing with us.
There are different types of AI, ranging from those that have become more popular since 2022, when we started having chat interactions. But many others involving various areas, and they are already present in our lives long before we realize it. For instance, there is an artificial intelligence (or several) that filters what kind of content we will receive in our social media feed.
There are generative artificial intelligences, which generate something, a text, an image, a video; but there are also others that, through various statistical logics, will bring certain content to us or make a forecast of how the weather will be. Artificial intelligence has been a part of our lives for much longer than we imagine, especially with computers and cell phones, but we've been talking about them even more recently due to the popularity of chat.
In early July, the United Nations (UN) held a meeting warning of the risks of artificial intelligence (AI) and advocating for accountability of developing companies. Considering the negative impacts that this technology may generate, what risks are already perceived in the democratic process, whether in Brazil or other parts of the world?
We are highly connected to our social media feeds, obviously, constantly, and this has been the case for a long time, not something recent. As I mentioned, there already exist artificial intelligences that guide what type of content we receive. And the goal of the major social media companies (big techs) is to make us spend as much time there as possible. In other words, the way artificial intelligence presents some content to us is designed to keep us there, to spend the longest possible time scrolling through the feed.
This already has a very strong impact on democracy, because there is a shift: we have changed the way we interact with the world, with people. Today, we receive a lot of content from people who think like us, who often have an interest in misinforming us. These echo chambers that exist within social media, or rather, these filters that make us receive the same content, are already shaped by artificial intelligences. We've been living this for a long time, even before 2022.
What leads to less deliberation, less effort to deal with differences and talk. Politically, this is a problem because democracy is precisely about dealing with not always the same thoughts, with the same people. We have diversity, and dissent is very important in democracy. Moreover, as a small piece of what we see as a reflection of artificial intelligences in democracy, we now have generative artificial intelligences.
When we think that today a person can be reproduced directly (whether candidate or fictional) to speak about certain ideas, this can reinforce prejudices, sync with hate speech, and generate misinformation about agendas that influence elections and democracy. This is very harmful.
Not that they cannot have beneficial uses; they do have beneficial uses. But something that really catches my attention is how today it's much easier to go into a chat and say: "Create content that resembles journalism but is misleading, something related to fake news." Or: "Create a character that resembles such a person, with such characteristics, who will say this." Thus, the balance between what is real and what is manufactured becomes more tenuous.
Currently, there is no legal regulation in force that addresses IAs in Brazil. The most advanced we have is PL 2.338/2023, authored by Senator Rodrigo Pacheco, which focuses primarily on the development, promotion, and ethical and responsible use of artificial intelligence in Brazil. What is your assessment of this regulation? What is its current status in terms of its progression, and how does it compare to regulations currently in place around the world?
It's very interesting that this bill exists in Brazil, it's very important. It is inspired by the AI Act, the European Union's legislative model, which has a logic of weighing the different types of AI based on the levels of risk they pose. There are uses that may pose more risks to our health, well-being, and democracy, and others that pose less risk. It's interesting this approach of thinking based on risks.
The reality is that there is a lot that could be improved in this PL. There is still a duality in the discussions that take place in Congress: either "we need to innovate and, therefore, require less regulation and less accountability from AI companies," or "we need to regulate more." It's as if there exists a paradox: we need to advance in AI because it is important for the economy, but we cannot do this with many restrictions, imposing many sanctions, because otherwise we will fall behind.
I think this line of thinking is very harmful to the advancement of AI, because there is an economic interest of large tech companies that this bill be approved in a way that holds them less accountable. It is very important that we have this discussion, taking into account that several other countries do not have a regulation or a bill to regulate AI. However, we see a very strong lobby from large tech companies within Congress, pursuing their own interests. We see more of an economic interest than a social one.
The large tech companies I am referring to here are foreign, typically from the United States, some from China, but mainly from the United States. AI is not the villain, it does not need to be seen as something malevolent, but the way we are currently developing the AI industry is heading in a concerning direction.
However, if we think about open model development, AI infrastructure, and models that are Brazilian, that will empower people here, and are used based on our data, that's when we're talking about AI benefits. But today, the way the PL is being developed and discussed is not, in my view, effective regarding this promising side of AI. So in practice, we risk having this bill greatly emptied and not fulfilling the function it proposes.
In this regulatory vacuum, today AI use in Brazil is governed by the General Data Protection Law. Specifically for elections, a resolution from the Superior Electoral Court (TSE) on electoral propaganda regulates AI use in this election. The Court and the General Advocacy of the Union (AGU) also signed a technical cooperation agreement to regulate and oversee AI use in election campaigns. So, what rules apply to the 2026 election regarding AI use, or what can and can't candidates do with these tools?
It's hard to say what can't be done because there are many legal and legal loopholes. We don't have extremely structured laws on the subject; we have some normative acts. What we can understand is: be very careful with the type of deepfake you use, as a candidate person.
We've seen the repercussions due to there being a person, who could not be displayed anywhere, not even through artificial intelligence, and whose image is being reproduced. The care that needs to be taken is: before the election period itself, any type of deepfake production or images presenting a candidate person must be totally halted, within 72 hours before, if I'm not mistaken.
But, for me, ethically, and in terms of impact, I believe this should not be used at all to represent someone who exists. Let's say: if I'm a candidate person, I'll use my own image, my own voice, and not an AI that will reproduce me. Because when we start thinking that candidates themselves are reproduced through AI, a loophole opens up to bring in other people, their competitors, into this game as well.
So, ethically, for me, it's very simple, it's okay, you can use generative artificial intelligence, as long as you're positioning it as such; as long as you're working with people who are created, who actually don't exist; and, moreover, that there's no hate speech or other issues that we already know are ethically interesting.
We still have a lot to advance in terms of what's allowed and what's not. The legal field is behind on how to deal with something that's forming now and that we still don't know how to handle in social life. Especially in elections, which is a totally contest context: ideological, political, economic, and of values.
Something problematic is that sometimes there's not a candidate person or their team using it, but voters producing content with the intent to generate some type of modification or friction in the elections. And this doesn't only apply to AI; it goes from decontextualizing content to working with misinformation. So, it's something much broader.
In your opinion, how does the debate about the use and regulation of AI relate to a broader, current discussion regarding our national sovereignty and foreign interference in our elections?
Today, it is virtually impossible to talk about national sovereignty without also discussing digital sovereignty. They are interconnected. We live in a digitalized, platformized society, which means that much of the technology we use is concentrated in the hands of a few companies. And few of them are Brazilian companies; most are American, some Chinese, some from other parts of Asia, but primarily American.
When we talk about how these foreign tech giants are in our lives, they are everywhere: from how public management works, or which are the large tech companies involved in Brazilian public management? Microsoft, Meta Group, Oracle, Amazon; we have many companies that hold our data, Brazilian data, data about us, citizens and citizens.
How is the ownership of this data by large companies used? What is done with this data? Even if there are contracts that say and large tech companies claim "not to use this data to train AI models," there is a history of breach of many contracts by foreign large tech companies.
We also see this use in a private context. Organizations do this as well, not just public or third-sector organizations; private organizations also use these technologies for many things. When I am on my social networks, data is collected there. When I am having a chat conversation with an AI, data is also collected. That is, there are many data points that can be directly used in different spheres, from data about an entire country or municipality to data about ordinary people that can be used for behavioral manipulation.
This is something that has already happened in the United States in the 2016 elections: various data was collected to understand the values, beliefs, and fears of people, and it was used to build messages, political propaganda, to make a particular candidate win and not another.
So, the possibilities of our data being used to reduce our national sovereignty, to increase geopolitical influence here, to manipulate our behaviors, or even to influence national elections, also exist.
Moreover, today, Brazil is a very targeted country by several foreign countries. We are also talking here about the United States and China. We have already seen that the US federal government has had interference postures in countries in Latin America, even at a political and direct level. And today, the United States is also targeting Brazil, including with political influence, so that those who interest them occupy positions in the presidency and Congress. This is to maintain the hegemony of the United States globally.
The issue is: we do not critically and frequently question what we are doing with our data. This is not just on an individual level, of course not; it is an issue that involves a structure as well. But we need national sovereignty, and for that, we need our own technological infrastructures, our AI infrastructures. We have to make sure our data stays here and not outside. We need to develop technologies that are ours, that are Brazilian, not just for public administration, but for everything else.
China has several stances that I believe are interesting, even though I also disagree with others. It has been an example at this level of national sovereignty allied with digital sovereignty: developing minds, people, technologies, and infrastructures for China, by China. We can and should draw inspiration from these positive initiatives that are also foreign.
Furthermore, how does the theme of AI fit into the geopolitical struggle for global hegemony, especially between China and the US? How do these two countries handle the issue?
So, to sustain the AI industry as it is today, with large tech companies concentrating much of the AI, there needs to be a lot of data. Thus, much data from the Global South goes to the Global North, mainly to the United States.
They also need labor, because, unlike many people think, to train a large AI model, not all, but these more generalist models (ChatGPT, GeminiI, Claude), need people to do moderation work. So, there are people there classifying whether an image is of a cell phone or not, evaluating responses. And these jobs are usually very low-paid and precarious. And typically, those doing this type of work are people from the Global South, including Brazil.
There is also a need for a lot of water and energy for these foreign tech companies to operate. Brazil is a country known for having "clean" energy, which does not mean there are no ecological risks.
There are several points to mention: there is a great interest from China and the United States in occupying other places in different ways, collecting data, and gaining access to the markets of some countries, through precarious work. And it's worth remembering, there are populations more susceptible to precarious work, like us here.
On another front, when we talk about critical metals, which exist in rare earths and help create the pieces and the infrastructure that supports AI models, Brazil is also a standout. Our territory has a lot of rare earth, the second-largest country after China in the world, according to the data known so far.
In other words, there is a desire to maintain hegemony over us, focusing primarily on the United States and imperialism, which seeks to take full advantage of other countries. And, obviously, today the United States is losing a bit of this hegemony due to technological advancements, particularly from China, in robotics and artificial intelligence. So, they are basically thinking: "Wow, for so long I was the empire, I dictated the rules, and now there's a country from Asia that wants to compete with me."
We see two very different stances in these two countries: while China seems to be turning inwards, developing its own artificial intelligences, its own technologies, but certainly also wanting to spread that to the rest of the world; in the United States, we see a more aggressive movement, much more focused on "I need to attack this growing enemy." Not a more national, internal investment for this to develop for the population, but for the maintenance of hegemony. We see very aggressive stances coming from the United States; not that this doesn't exist to some extent in China, but it's much inferior.
In other words, we are seeing this competition, with different strategies. But we are in globalization, so it's impossible to think of full independence, economically and even politically. This friction will continue and it will generate many military issues, because AI is also being fed today for military purposes, which is another sad reality.
Often, we think: "Wow, people are just creating cute cat videos, singing, right?" No, AI is also being used for military purposes, in autonomous weapons that choose targets, including for data collection. Someone can invade a country, collect lots of data from that country while destroying it for war purposes.
On the other hand, even if they are not neutral, AI is not inherently a bad technology. So, what positive applications can come from artificial intelligences?
We already have many very positive applications in many areas, starting with medicine. Not all applications in this area are necessarily positive, but today there are several AIs that can process a very, very large number of data points. Something we, as individuals, cannot do. For diagnoses or identifying a cell that may have suffered a mutation, this is already being applied in medicine and it's something positive.
We also see that digital sovereignty is not only restricted to what we understand as national. A specific population from a certain place can develop its own artificial intelligence to solve problems or find opportunities within that context.
Let's imagine: we are a small community that lives off the land and is having soil problems. How can there be the development of an artificial intelligence to solve some kind of problem there, to see what's causing possible (let's say) erosion? What can be done is to identify soil patterns, identify rainfall patterns, identify air humidity; there are many possibilities.
Something very interesting we can do at a more individual level is that there are some AI models you can install on your own computer. So, you won't need to use an AI linked to a brand that collects your data.
Let's say I want to use here an open-source AI, meaning one that allows me to understand how it works more precisely because it's not under commercial secrecy, and my data isn't necessarily being collected for some intent. So, I can set it up on my computer and, with that, optimize my routine and take care of my small business, or think about optimizing my production routines.
That is, there are alternatives to the general large AI models that society is using a lot. There is an infinity of possibilities to make us reimagine what artificial intelligence is. Today, our imagination and what we understand by artificial intelligence is very restricted to what foreign tech giants are offering to us.
This is very bad because we are losing the possibility of imagining other possible futures from our needs. And this requires investment in technology, information, and our infrastructure. So that all people, at a minimum, understand how an AI works. And I'm not saying they need to deeply understand programming languages, but there should be ways to make the general population digitally literate and receive a certain type of knowledge, so they can identify how they can use that knowledge to their local context. We don't need an AI to solve everything.
So-called deepfakes should be among the main challenges of the digital environment in the 2026 elections. How can the average voter identify and protect themselves from deepfakes or other AI-generated misinformation on social media?
It's very common for me, who studies this, to often watch a video and not know if it was made by AI or not. There are some more advanced computing techniques that will try to understand if something was made with AI or not. But if I don't have access to them, what do I do?
Then we need to think less about technology and more about critical reflection. What I'm seeing seems to be confirming something I believe? Is it something that seems very opportune to appear in the electoral context or in the context of ideological, political dispute? We need to roll up our sleeves and check.
I need to see: is there any fact-checking institution that has weighed if it was manufactured or not? Suddenly, there might be some information in the comments; someone appears saying: "This is misinformation, this was manufactured, here's the link that explains the whole process."
We need to have this reflection, which requires us to step away from the accelerated logic of content consumption. Short videos are already speeding up our thinking logic. We receive a lot of information in a short time. We have an enormous saturation of content in our lives. And often we lose criticality, our ability to reflect; then we assume that everything we're seeing is reality.
What we cannot lose perspective on is: maybe you really don't understand if that video is manufactured just by looking at it. You might need to seek other information to verify. So, what I invite first is: analyze, as much as possible, some things in the video.
If, at any moment, a hand appears out of nowhere, or a face with a certain pattern suddenly changes, or you see a very quick movement that seems to blur the details, because when you go slower you can see more details. But, of course, we already have artificial intelligences that can simulate the highest realism.
But the second step is: doubt, research, and inform yourself. Be very careful with confirmation bias, thinking: "Look, this makes a lot of sense," because it confirms a belief of yours. I am a bit obsessive about this: I already doubt many things.
Often it's real, but I will check if that information came out, not necessarily in major outlets, but in journalistic vehicles that I already know have credibility. Maybe it wasn't a person I've never seen in my life who posted the video; it was a journalism vehicle that I believe conducted a serious investigation and understands if that's real or not. So we need criticality, critical reflection.