Brazilians send approximately 140 million messages per day to ChatGPT. This figure, disclosed by OpenAI itself, places Brazil as the third-largest user of the tool in the world, behind only the United States and India. We are one of the societies that has most widely adopted artificial intelligence on the planet and, at the same time, one of the countries that still lacks a law to regulate it.

There is a paradox surrounding artificial intelligence policy in Brazil. The country that produced one of the world’s most advanced digital electoral regulations, approved an AI legal framework bill (Bill 2,338/2023) in the Federal Senate, and launched a national plan with BRL 23 billion in investment still sees deepfakes of political leaders circulating during an election year. In other words, the law exists on paper while technology advances in the real world. And it is precisely in this gap between the two worlds that the most urgent debate for our democracy resides.

The issue is not simple. AI regulation in Brazil simultaneously involves technological sovereignty, fundamental rights, algorithmic bias, democratic protection, and economic competitiveness.

The Brazilian AI Plan: ambition equal to the challenge?

In July 2024, during the 5th National Conference on Science, Technology, and Innovation, the federal government presented President Lula with the Brazilian Artificial Intelligence Plan (PBIA 2024–2028), titled “AI for the Good of All.” The name reveals a political definition: in Brazil, responsible AI will not be treated merely as an agenda for economic competitiveness, but also as an instrument of social well-being.

The PBIA is structured around five strategic pillars, 54 actions, and an estimated investment of BRL 23 billion. Its core objectives include developing large-scale language models (LLMs) in Portuguese, based on national data — a direct bet on technological and linguistic sovereignty — as well as creating a supercomputer ranked among the five most powerful in the world to expand Brazil’s national research and processing capacity.

The logic of the plan is to leverage Brazil’s real comparative advantages: a predominantly clean energy matrix, which is essential for sustainable data centers; cutting-edge scientific research capacity; and consolidated technological expertise in strategic sectors such as agriculture, healthcare, and the environment.

The question that remains on the mind of anyone following this agenda, however, is always the same: between announcement and execution, how many Brazilian plans have been lost along the way? The PBIA has structure, resources, and political visibility. What it still needs to prove is whether it has sufficient governance to move beyond paper.

The AI legal framework: five years of debate, a law that has yet to arrive

AI regulation in Brazil has a long history for such a recent technology. Since 2019, at least six bills addressing the topic have moved through Congress. In 2022, the President of the Senate, Rodrigo Pacheco, appointed a Committee of Jurists to develop a robust proposal. The result was Bill 2,338/2023 — the so-called AI legal framework — which, after more than a year of public hearings, sectoral debates, and political negotiations, was unanimously approved by the Federal Senate on December 10, 2024.

The text represents a paradigm shift: AI is no longer treated solely as a technological tool and is instead regulated through a logic of risk, impact, and accountability. This means that there is no single rule for every AI application, and that the regulatory burden must be calibrated according to the potential harm of each context.

In other words, a system that recommends songs on a streaming platform and one that decides whether credit should be granted or whether a defendant should receive provisional release cannot be treated in the same way. This logic is directly inspired by the European AI Act, a global reference that entered into force in 2024 and influenced the Brazilian debate by consolidating the risk-based approach as the central axis of regulation.

It is also worth noting the significance of the AI Act for Brazilian companies operating in Europe: in August 2026, its sanctions will become fully applicable, with fines of up to €35 million or 7% of global turnover for those who fail to comply with the rules governing high-risk systems.

In the case of Brazil’s AI legal framework, the central points include: the prohibition of systems that induce manipulative behavior or encourage discrimination; the prohibition of AI-generated child sexual abuse content; the protection of copyrighted works; and requirements for transparency and the right to contest algorithmic decisions when they affect people’s rights. The bill provides for penalties of up to BRL 50 million per violation.

Approved by the Senate, the bill was sent to the Chamber of Deputies in March 2025, but it arrived with a problem. The government itself identified a constitutional flaw in the text: parts that should have been proposed by the Executive Branch had been included by the Legislative Branch, which could cause the law to be invalidated by the Federal Supreme Court. To correct this, in December 2025 the government sent Congress a complementary bill creating the National AI Governance System (SIA).

Therefore, in practice, two bills are moving forward together to form a single law. Meanwhile, pressure from the private sector, disputes over who will oversee AI in the country, and the 2026 electoral calendar are making final approval increasingly uncertain.

Brazil does not just need to regulate AI. It needs to create it.

Brasília, November 4–6, 2026. The gathering that begins to change that.

Discover the event

Algorithmic bias and responsible AI: what the law must guarantee

Much of the debate on AI regulation in Brazil focuses on tangible issues such as deepfakes, disinformation, and job automation. But there is a layer that is more difficult to take into account — and equally dangerous — that must be at the center of the legal framework: algorithmic bias.

AI systems learn from historical data that carry the historical inequalities of those who produced them. In other words, an algorithm trained on Brazilian credit data tends to reproduce the patterns of financial exclusion that have always penalized Black people, women, and residents of peripheral communities. A facial recognition system trained predominantly on white faces systematically misidentifies Black people, even leading to cases of wrongful arrests in other countries.

Responsible AI is not a slogan. It is a set of practices that must be codified into law and overseen by an agency with the resources and independence to act. These practices include model audits, diversity in training data, transparency in decision-making criteria, and real mechanisms for contestation.

This is why the discussion about Brazil’s AI legal framework cannot be limited to prohibitions. It must also answer complex questions: Who audits the algorithms that decide who receives social benefits, who passes through employment screening, or who is flagged as a “risk” in public security systems?

AI and disinformation: the threat that does not wait for the law

While the AI legal framework moves through the corridors of Congress, disinformation amplified by artificial intelligence is advancing at a speed incompatible with the pace of legislation.

Between 2024 and 2025, the number of deepfakes and false AI-generated content in Agência Lupa’s fact-checks jumped from 39 to 159 cases — an increase of more than 300%. In 2025, this content accounted for 25% of all verifications conducted by the agency. More than three-quarters of these materials exploited the image or voice of public figures, especially political leaders.

The role of the Legislature: what already exists and what is still missing

It is easy — and to a large extent fair — to criticize Congress for the slow pace of the debate. But the full picture involves more nuance.

Brazilian electoral legislation has advanced on concrete digital fronts, such as Resolution TSE 23,732/2024, which prohibits the use of deepfakes in electoral campaigns. In addition, there is already an obligation to disclose when AI is used in electoral advertising, along with restrictions on the use of bots to mediate contact with voters and platform liability for content containing disinformation and anti-democratic material. Researchers who monitor the global landscape recognize that, in terms of digital electoral regulation, Brazil is among the most advanced countries in the world.

The process of drafting the bill aimed at establishing the AI Legal Framework (Bill 2,338/2023) also advanced with the participation of more than 300 people, including experts, representatives of the private sector, civil society, and oversight bodies, in addition to 117 represented institutions. Its approval brought together senators from different ideological orientations around a text built by consensus.

The problem is speed. The technology that the Legislature seeks to regulate evolves faster than any democratic process can keep up with. And the gap between the approval of the law and its effective implementation — with a structured regulatory body, budget, and real enforcement capacity — tends to be even longer than the legislative debate itself.

There is also a dispute of visions that the text of the law alone does not resolve: on one side, there are those who argue that excessively restrictive AI regulation in Brazil will stifle startups and place the country at a global competitive disadvantage. On the other, there are those who argue that the absence of regulation gives excessive power to technology giants without requiring accountability to the rest of society.

Far beyond technology

The discussion about Artificial Intelligence is not only about software, algorithms, or infrastructure.

It involves economic development, competitiveness, democracy, technological sovereignty, and innovation capacity.

The future of Artificial Intelligence will be defined not only by those who create the technology, but also by those who build the institutions capable of guiding its use.

And this is one of the most important conversations Brazil needs to have at this moment.

This is one of the discussions in the Transformative AI track of the FUTURA AI Conference

Between November 4 and 6, international researchers, representatives of government, the Judiciary, the private sector, and academia will gather in Brasília to discuss the future of Artificial Intelligence in Brazil and around the world.

Because the question is no longer whether Artificial Intelligence will transform society. The question is how we want to participate in that transformation.

Join the FUTURA AI Conference and closely follow the discussions that are helping to build the future of Brazilian Artificial Intelligence.