The preliminary text of Bill 2338/2023 (PL 2338/23), known as the Artificial Intelligence (AI) Legal Framework, is currently being reviewed by the Internal Temporary Committee on Artificial Intelligence in Brazil (CTIA) of the Federal Senate.

Legislative discussions began in 2020, facing controversy over the need for a specific regulation for AI. There were other bills on the subject in both the Chamber of Deputies and the Federal Senate, which were incorporated into Bill 2338/23 by a legislative agreement.

The text of Bill 2338/23 defines “artificial intelligence system” as a machine-based system that uses data or information to generate forecasts, recommendations, or decisions, and can influence the virtual or real environment, including generative AI. 

The Brazilian bill is aligned with international practices, as it also adopts a risk classification system, similar to the European AI Act and to the U.S. AI Executive Order, guaranteeing the rights of people affected by AI. 

These rights include the right to prior information on interactions with AI, personal data privacy and protection, human participation in AI system decisions, and non-discrimination. 

The rules for regulating AI set out in the text of the bill do not apply to systems developed for private, non-economic use, national defense, research activities, or those in open and free formats, unless they are considered high risk. 

On the other hand, the bill prohibits the use of AI that induces unhealthy behavior, autonomous weapons systems without substantial human control, and AI to assess criminal risks or rank individuals universally. Systems that facilitate the creation or dissemination of child sexual abuse material are also prohibited. 

The bill also addresses the protection of copyright content and defines fair use for research institutions, journalism, museums, archives, libraries, and other educational entities. 2 

In addition, it proposes the creation of the National Artificial Intelligence Regulation and Governance System (SIA), a government authority in charge of establishing rules, overseeing the application of AI-related rules, and imposing the penalties provided for. 

Among the possible punishments, the bill provides for warnings and fines of up to BRL 50 million per violation, or up to 2% of the company’s sales revenue. There could also be a ban on participation in regulatory sandbox regimes for up to five years and suspension of AI-related activities. 

The CTIA was due to present the final report on the wording of Bill 2338/23 last June 18. However, after the incorporation of several last-minute amendments related to topics such as risk assessment, governance, labor relations, protection of consumer rights, self-regulation, good practices, inspection, and sustainable innovation, the senators decided that it would be necessary to extend the term for analyzing the report, as well as to hold another five public hearings. 

* * * * * * * * 

For more information on the above or other matters, please contact Maristela SA Rossetti (mar@rraa.com.br) or Gilberto Rossetti (gmr@rraa.com.br). 

This article is based on publicly available information and given for informational purposes only. It is not intended as legal advice foreign subsidiary as a comprehensive analysis of the matters referred to herein.