Brazil: Brazilian AI Regulation Bill No. 2,338/2023

1. Overview

On December 10, 2024, Brazil’s Senate approved Bill No. 2,338/2023, which outlines sweeping provisions relating to artificial intelligence (AI) and, while neither targeted at nor addressing deepfakes, aims to foster the protection of human dignity and rights, promote the prevention of misuse of AI systems, and increase transparency on the use of AI technologies and their outputs concerning content generated by AI, purportedly including deepfakes. 

2. Application of the law

It applies to any AI system that poses risks to human rights, as well as to systems already able to create various types of manipulative content like deepfakes. It covers the creation and dissemination of content. Systems that are privately designed for individual use and not for economic activity, defense or for purposes of research or to model open and free forms (with the exception of systems designed as high-risk) are excluded.

3. Key provisions

• Prohibited behavior: The Act is against the use of AI systems that utilize subliminal means targeting or intended to induce a natural person to act in a dangerous or dangerous to their health or safety manner.
• Consents Requirement: Specifically, people must be Informed when interacting with an AI system, given explanations about, or otherwise information relating to, the decisions of an AI system, capable of appealing and requesting human review of an AI system’s decisions.
• Platform Responsibilities: AI providers must carry out first-time risk assessments, implement governance models, i.e., transparency and security standards, and in high-risk systems, maintain technical documentation, log registers, and conduct reliability tests.

4. Penalties & Enforcement

Penalties:

•  Up to BRL 50 million per offense or up to 2% of the company’s revenue from sales.
• Suspension of AI-related activities.
• Exclusion from regulatory sandbox regimes for a term of up to five years.

Enforcement Authority:The National Artificial Intelligence Regulation and Governance System (SIA) is established to serve the aim of regulating the use of AI, with the coordination of the regulatory process by the National Data Protection Authority (ANPD).

5. Notable Cases or Precedents

There have been no publicly known court cases testing this law to date.

6. Comparison to Global Standards

Brazil’s approach mirrors the European Union’s AI Act in sorting AI systems by risk and applying tough regulations to high-risk applications. Certain safeguards, particularly to ensure content integrity and political processes, were cut from the final draft.

7. Real-World Implications

Developers and platforms of AI must adhere to risk evaluations and governance requirements. People should be aware of their rights to information, explanation, and contesting AI-driven decisions.

8. Future Prospect

The bill is under consideration by the Chamber of Deputies, awaiting further discussion. The debates continuing may lead to amendments, especially on controlling generative AI and deepfake technologies.