- Brazil’s Senate passed AI Bill PL 2338/2023 in December 2024, establishing a risk-based framework that prohibits “excessive risk” AI systems and imposes strict rules on high-risk applications.
- The bill moved to the Chamber of Deputies in March 2025, where a special committee is analyzing it amid contentious debates over copyright provisions and high-risk classifications.
- PL 2338 requires compensation for AI training data derived from copyrighted works — a stricter stance than the EU or Japan, which provide broader exceptions.
- Final passage depends on resolving congressional disagreements before Brazil’s October 2026 general elections, which create a practical legislative deadline.
What Happened
In December 2024, the Brazilian Federal Senate approved PL 2338/2023, a comprehensive bill to regulate artificial intelligence in Brazil. Originally proposed by Senator Rodrigo Pacheco (PSD/MG) and developed by a Committee of Jurists appointed by the legislative branch, the bill underwent extensive revision under the rapporteurship of Senator Eduardo Gomes (PL/TO) throughout 2024.
On March 17, 2025, the Senate formally forwarded the approved bill to the Chamber of Deputies for review. On April 29, 2025, the Chamber’s president established a special committee to analyze the legislation, and public hearings ran through late 2025 with testimony from technology companies, civil society organizations, and legal scholars.
Why It Matters
PL 2338 would make Brazil the first country in Latin America to adopt a comprehensive AI regulatory framework. The bill follows a risk-based approach structurally similar to the EU AI Act, but includes provisions on copyright and training data that go further than any other major jurisdiction currently in force.
According to the Library of Congress’s analysis, the bill categorizes AI systems into three tiers: excessive risk (prohibited outright), high risk (subject to strict compliance obligations), and general use (lighter transparency requirements). The framework covers AI deployed in sectors including healthcare, education, hiring, autonomous vehicles, and biometric identification.
Brazil’s competition authority CADE has suggested amendments to the bill, reflecting ongoing institutional concerns about its potential impact on market competition and the country’s growing AI sector.
Technical Details
The bill prohibits “excessive risk” AI systems outright — specifically those using subliminal techniques to manipulate human behavior or that exploit the vulnerabilities of specific groups such as children, the elderly, or people with disabilities. High-risk AI systems face mandatory requirements including impact assessments before deployment, human oversight mechanisms, transparency obligations toward affected individuals, and detailed documentation of training data sources and methodologies.
The copyright provisions have proven particularly contentious in the Chamber of Deputies. PL 2338 requires companies to compensate rights holders for copyrighted content used to train AI models. A limited text and data mining exception applies to non-commercial research institutions, journalism organizations, museums, archives, libraries, and educational entities, but explicitly does not extend to commercial actors or for-profit companies. This is stricter than both the EU AI Act and Japan’s copyright framework, which provide broader fair-use exceptions for AI training purposes.
The bill establishes a national AI authority responsible for enforcement, registration of high-risk systems, and development of sector-specific guidance, though details about the authority’s institutional structure, staffing, and budget remain under active discussion. Penalties for non-compliance include monetary fines, temporary or permanent suspension of AI system operations, and mandatory public disclosure of violations.
Who’s Affected
All companies developing or deploying AI systems in Brazil would fall under the framework’s scope, with obligations scaling according to risk classification. International AI providers — including large language model developers and cloud platforms — face particular exposure on the copyright provisions, which could require licensing agreements or direct compensation for training data sourced from Brazilian content creators and publishers.
The creative industry, represented by organizations such as CISAC, has strongly supported the compensation requirements. Technology companies and industry groups have argued that the provisions could deter AI investment in Brazil and push development to jurisdictions with more permissive intellectual property rules.
What’s Next
The Chamber of Deputies’ special committee continues deliberations, with a floor vote possible in early-to-mid 2026. If the Chamber amends the bill in any substantive way, the revised text must return to the Senate for final approval before reaching President Lula for signature. Brazil’s October 2026 general elections create a hard practical deadline — legislation not passed before the current congressional session ends risks being shelved indefinitely, requiring reintroduction in a new Congress with potentially different political priorities.
Related Reading
- White House Releases Federal AI Framework, Urges Congress Against Heavy Regulation
- UK AI Policy: Pro-Innovation Approach With AI Bill Expected in 2026
- Singapore Leads on Agentic AI Governance With World-First Framework
- LlamaIndex Review 2026: Data Framework for Building RAG and Knowledge Applications
- China AI Regulation: The World’s Most Layered AI Governance Framework
