Brazilian Bill regulating the use of Artificial Intelligence is approved by Senate and goes to the Chamber of Deputies

By Natali de Vicente Santos Kapulskis e Sarah Carneiro

After more than a year of debates, the Senate yesterday (10) approved Bill No. 2.338/2023, which aims to regulate the use of artificial intelligence, proposed by Senator and President of the National Congress Rodrigo Pacheco (PSD-MG), under the terms of the substitute text presented by its rapporteur, Senator Eduardo Gomes (PL-TO).

The Bill, already analysed by CGM Advogados in a previous post (link here), was approved by consensus by the Internal Temporary Committee on Artificial Intelligence in Brazil (“CTIA”) last Thursday (05), after a complex process that involved analysing at least 86 amendments. Of these, only 27 were entirely approved.

Among the highlights of the approved Bill, which creates the National Artificial Intelligence Regulation and Governance System, we mention:

• Degree of risk: the proposal provides that system developers, suppliers or applicators may, or must (in the case of generative and general-purpose systems), carry out a preliminary risk assessment, based on the purposes and impacts of AI systems on human life and fundamental rights.
• ANPD, regulation and penalties: the National Data Protection Agency (“ANPD”) has been envisaged as the body responsible for imposing fines of up to BRL 50 million, or 2% of the gross turnover of companies that violate the rules set out in the Bill. In addition, the ANPD will also be responsible for issuing rules on certification forms and requirements, algorithmic impact assessment procedures and for reporting serious incidents.
• Copyright: the owners of works used for training new systems must be compensated.
• Individual guarantees: (i) for systems that make biometric identification, protection against direct, indirect, illegal or abusive discrimination will be guaranteed; (ii) jurisdictions will have the right to an explanation and human review of decisions that have a relevant legal impact; and (iii) the development of training and continuous qualification programs will be encouraged, promoting the appreciation and professional improvement of workers; and
• Freedom of expression and integrity of information: a section has been deleted that linked the integrity of information to the responsibility of generative AI, as this could act as a mechanism for controlling freedom of expression and access to information.

The Bill also provides for a list of artificial intelligence systems considered to be high risk and therefore subject to stricter regulation, which includes the following, as mentioned on the Federal Senate’s website (link here):

• autonomous vehicles.
• traffic control and management of water and electricity supplies when there is a danger to people’s physical integrity or a risk of unlawful or abusive interruption of services.
• selection of students for access to education and academic progression.
• making decisions on recruitment, evaluation, promotion and dismissal of workers.
• evaluating criteria to assess eligibility for services and public policies.
• investigating facts and applying the law when there are risks to individual freedoms, within the framework of the administration of justice.
• Priority management in emergency services, such as firefighters and medical assistance.
• analytical study of crimes.
• medical diagnostics.
• border control.
• emotion recognition by biometric identification; and
• data analysis for crime prevention.

The Bill has been submitted to the Chamber of Deputies for analysis and approval. If approved, the bill will go to the President of the Republic for sanction and then become law. We will continue to follow the discussions on this matter and its practical developments.