Special Commission that analyzes PL 2,338 of 2023 promoted debate on the subject on Tuesday (12.ago)
Artificial Intelligence Experts heard on Tuesday (12.ago.2025) in the House of Representatives defended flexible and decentralized regulatory model for the sector. Participated in the debate of the Special Commission that analyzes the PL 2,338 of 2023, on governance and regulation of AI systems.
According to them, the project under analysis in the House is more aligned with the European model and needs adjustments to avoid what they called “Disproportionate regulatory load”. This approach would have negative impacts on the competitive efficiency of national companies, especially startups and small businesses as the costs of innovation increases.
Currently, 2 regulatory AI models stand out: the most prescriptive and centralized, adopted by the European Union (IA Act), and another closer to self -regulation and decentralization, in force in the United States.
In the case of Brazil, the director of digital transformation of the Ministry of Development, Industry, Commerce and Services, Cristiane Rauen, bets on a “Tropicalized Model” adapted to domestic market conditions.
For the director, the ideal would be to avoid the “compliance ex ante”when the rules are imposed before the use of AI, which, in your view, increases bureaucracy and makes innovation difficult for small businesses, as well as being little responsive to technological changes.
“For us, it is interesting to have a balance model, which is not very punitive from the point of view of innovation. A asymmetrical regulation model, ie the higher the risk of AI, the higher the regulatory load and the greater the duties that have to be applied.”defended.
At this point, the rapporteur of the proposal, Deputy Aguinaldo Ribeiro (PP-PB), said that the definition of which regulatory model will be adopted must be a state decision and not political parties, based on the sovereignty of the country. “If we do not have this, our data will not be about our control”he declared.
SIGHY INFORMATION
“It is not bringing model copied from Europe that will work well in Brazil”said Jean Paul Torres Neumann, executive of Assespro (confederation of associations of Brazilian information technology companies), which represents the IT companies affected by the bill.
Project critic, Neumann said the text can undermine industry competitiveness by exposing strategic business information. “The company is no longer competitive for disseminating information that is confidential”he said. He referred to the project article by which AI developers must share high -risk algorithmic impact assessments with the system’s sectoral authorities.
Under the proposal, it will be up to the competent authority by the system, the ANPD (National Data Protection Agency), the decision on the general criteria for the algorithmic impact assessment. The sectoral authority will be responsible for defining in which hypotheses the assessment of algorithmic impact will be flexible. The project does not explain who decides in cases of differences.
As suggestions to the collegiate, Neumann proposed the creation of credit lines and Specific AI tax incentives, similar to those foreseen in. In addition, it defended inclusion in the regulatory framework of supercomputer sharing as a cost reduction measure.
Excessive rules
The representative of CNI (National Confederation of Industry), Rodrigo Pastil Pontes, classified the proposal of the regulatory framework of “Excessive” to regular from the conception and development of technology to use. For him, the text “Breaks the development of innovation”harming competitiveness.
Bridges argues that AI regulation emphasizes high -risk uses rather than technology itself, excluding low and medium risk applications that do not use human data.
By the proposal under analysis in the House, the ANPD, as “Competent authority” By system management, it will be responsible for regulating the classification of the list of AI systems evaluated as high risk. This centrality of decisions at the agency was one of the most criticized points of the project.
Data Power
In the assessment of director Cristiane Rauen, Brazil is a power in terms of data that can be used to improve AI. According to her, there are 175 million users registered in Pix and 169 million active accounts on the site gov.br.
“It is a giant opportunity for Brazil to position itself in relation to the data: what we want from this data and how this data can be better benefited within AI legislation that does not focus only on the defense of rights and the positive compliance, but that associates these two ends with the promotion of the sector.”said Cristiane.
With information from .
