Quick Answer:
Brazil still has no comprehensive AI law. The Senate approved Bill PL 2338/2023 on December 10, 2024, but it remains pending in the Chamber of Deputies. Copyright treatment of AI training data and how high-risk systems are classified remain unresolved, and the second half of 2026 will pivot to election campaigning, according to Tech Policy Press.
Key Takeaways:
- Senate approval: the Senate approved PL 2338/2023 on December 10, 2024, but the bill still needs Chamber of Deputies approval and a presidential signature before becoming law, according to the Artificial Intelligence Act tracker.
- Two unresolved fights: copyright treatment of AI training data and the definition of high-risk AI systems remain the most contested issues, according to Tech Policy Press.
- Electoral pressure: the initial Chamber vote, once expected before the end of 2025, was postponed to February 2026, and the second half of the year will be dominated by campaign season, according to Tech Policy Press.
- Broad supervisory structure: the bill creates a National System for AI Regulation and Governance coordinated by the National Data Protection Authority (ANPD) alongside sectoral regulators including Anatel, Anvisa, Anac, Aneel, and the Central Bank, according to Data Privacy Brasil.
- Real penalties: violations could carry fines of up to BRL 50 million (about $1.6 million) or 2% of a company's annual revenue in Brazil, according to the Artificial Intelligence Act tracker.
Brazil's Congress is approaching one of the most decisive moments for its artificial intelligence bill since the Senate approved it more than a year ago. PL 2338/2023 is still waiting its turn in the Chamber of Deputies, and according to Tech Policy Press, the initial vote once expected before the end of 2025 was postponed to February 2026 — leaving an increasingly narrow window before the country's legislative calendar pivots toward the elections.
If passed, PL 2338/2023 would make Brazil the first country in Latin America with a comprehensive, horizontal artificial intelligence regulatory framework. If it is not passed before Congress's attention shifts fully to campaigning, the bill could sit in indefinite limbo while lawmakers focus on their own re-election bids.
What the Senate Already Approved
The Federal Senate approved PL 2338/2023 on December 10, 2024, as confirmed by the Artificial Intelligence Act tracker. The text was originally advanced by Senator Rodrigo Pacheco (PSD/MG) after a year of work by a Committee of Jurists, and the Senate's Internal Temporary AI Committee (CTIA) approved the substitute report on December 5, 2024, with Senator Eduardo Gomes (PL/TO) serving as rapporteur, according to Data Privacy Brasil's analysis.
The approved text establishes a tiered approach based on risk level. According to Data Privacy Brasil, it prohibits "excessive-risk" systems — such as autonomous weapons, real-time remote facial recognition (with limited exceptions), and systems targeting vulnerable populations — while imposing stricter obligations on "high-risk" systems, including those used in critical infrastructure, education selection, employment decisions, access to essential services, autonomous vehicles, medical diagnosis, and biometric emotion recognition.
Rights the text grants to individuals: according to Data Privacy Brasil, the bill grants the right to know when interacting with an AI system, the right to privacy and personal data protection, the right to non-discrimination, the right to an explanation of high-risk system decisions, the right to contest those decisions, and the right to human review.
On transparency, the text requires a simplified preliminary self-assessment of risk level for all systems and a mandatory algorithmic impact assessment before any high-risk system reaches the market, according to Data Privacy Brasil. AI-generated synthetic content would need to carry an identifier that verifies its origin.
The Two Fights Still Holding Up the Bill
According to Tech Policy Press, two issues account for most of the legislative controversy: the copyright treatment of data used to train AI models, and how a "high-risk" system gets defined and classified. Neither has been definitively settled since the Senate approved its version of the text.
On copyright, Data Privacy Brasil details that the Senate text requires developers to disclose which protected content they used to train their models, lets rights holders prohibit uses that contradict their interests as creators, and establishes a fair compensation system for copyright holders, with exemptions for non-commercial research by scientific institutions. The Artificial Intelligence Act tracker further describes an exception that would let developers use copyrighted material for training when it was obtained for non-profit purposes.
The classification of high-risk systems remains the other point of friction. The broader that category gets, the more companies — and the more types of AI-powered products — fall under mandatory impact assessments, disclosure requirements, and heightened oversight before reaching the market.
Why the Election Calendar Changes the Urgency
According to Tech Policy Press, the second half of 2026 in Brazil is fully dedicated to campaigning, which sharply narrows the window available for the Chamber of Deputies to pass the text. The same analysis notes there is a legitimate interest in seeing the regulation move forward before the elections, given that unregulated use of AI could directly affect the electoral process, and that Brazil's existing electoral court resolutions are seen as insufficient to cover the full range of risks posed by generative AI during a campaign.
Tech Policy Press also reports that the debate has been complicated by issues that emerged after the original bill was introduced, including data center regulation, and that the executive branch has proposed parallel regulations — including data center rules and a digital markets bill — competing for space on the same legislative agenda.
A Broad Coalition, and an Organized Opposition
The bill does not arrive at this stage without support. According to Data Privacy Brasil, organizations including Febraban (Brazil's banking federation), CNSaúde, and FIESP, along with the Coalition for Digital Rights, backed the text's advancement, and more than 35,000 citizens participated through the National Congress's e-Cidadania portal. A public letter signed by figures including Fernanda Torres, Chico Buarque, Pedro Bial, Milton Nascimento, and Caetano Veloso also backed stronger protections for content creators against AI training.
On the other side, Data Privacy Brasil documents that the National Confederation of Industries (CNI) pushed to dilute the original text's worker protections, and that federal-government-backed amendments weakened information integrity provisions. In the Senate's final text, information integrity was removed as a risk criterion under Article 15, generative AI systemic risk provisions were dropped, a new Article 77 was added stating that content moderation regulation requires separate legislation, and worker protections — including mass layoff provisions and worker participation in impact assessments — were weakened.
How Compliance Would Be Supervised
Unlike the European Union's centralized model, Brazil opted to lean on its existing sectoral regulators' capacity. According to Data Privacy Brasil, the bill creates a National System for AI Regulation and Governance in which the National Data Protection Authority (ANPD) acts as the coordinating body, alongside sectoral authorities such as Anatel, Anvisa, Anac, Aneel, and the Central Bank, a Permanent Regulatory Cooperation Council (CRIA), and a Committee of AI Experts and Scientists (CECIA).
The text also includes incentives: a regulatory sandbox for testing AI systems in a limited-period experimental environment, priority access to those testing environments for startups, and sustainability guidelines aimed at energy efficiency and environmental impact certification for data centers, according to the same analysis.
The penalties at stake:
- Up to BRL 50 million (about $1.6 million) per infraction, according to the Artificial Intelligence Act tracker
- Or 2% of the offending company's total annual revenue in Brazil, whichever is greater
What This Means for Your Business
While PL 2338/2023 applies directly to Brazil, the signal it sends is broader: AI systems that interact with the public — from chatbots to auto-generated content — are moving toward a standard of transparency and verifiability, not opacity. The bill's own text requires synthetic content to carry a verifiable identifier of its origin. That is exactly the terrain where Answer Engine Optimization (AEO) competes: making sure the information about your business online is structured, verifiable, and consistent, so AI tools — ChatGPT, Perplexity, Google AI Overviews — can confidently cite you instead of a competitor.
If your business depends on customers finding you through search or AI assistants, the real question is not whether AI regulation advances in Brazil or anywhere else — it is whether your own digital presence already meets the transparency and data-structure standard that AI needs to cite you with confidence. MerchandisePROS's free audit checks exactly that: it verifies your schema.org markup, your FAQ content, and your authoritative citations, and delivers a 0-to-100 score in under 60 seconds.
Frequently Asked Questions
What is Brazil's PL 2338/2023 bill?
PL 2338/2023 is the bill that would establish Brazil's first comprehensive artificial intelligence regulation framework, built around a risk-based approach, transparency obligations, and a national supervisory structure. The Senate approved it on December 10, 2024, but it still needs Chamber of Deputies approval and a presidential signature before becoming law.
Is AI regulation already law in Brazil?
No. The Federal Senate approved PL 2338/2023 on December 10, 2024, but the bill remains pending in the Chamber of Deputies. According to Tech Policy Press, an initial vote once expected before the end of 2025 was postponed to February 2026, so the law is not yet in effect.
What are the two most contested issues in the bill?
According to Tech Policy Press, the two most contentious issues are the copyright treatment of data used to train AI models and how systems get classified as high-risk.
What penalties does the bill set out?
According to the Artificial Intelligence Act tracker, violations could carry fines of up to BRL 50 million (about $1.6 million) or 2% of the company's annual revenue in Brazil.
Who would oversee compliance with Brazil's AI law?
According to Data Privacy Brasil, the bill creates a National System for AI Regulation and Governance (SIA) coordinated by the National Data Protection Authority (ANPD), alongside sectoral regulators including Anatel, Anvisa, Anac, Aneel, and the Central Bank, rather than a single centralized regulator like the EU's.
"Regulation is coming, in Brazil and across the region. Businesses that are already transparent and verifiable to AI will not have to scramble — they will already be on the right side of the standard."
- Diego Medina F, Founder of MerchandisePROS