In the continuation of our series on competition and regulation of digital markets and artificial intelligence, we analyze the latest highlights involving the two substitute proposals presented by Senator Eduardo Gomes (PL-TO) within the scope of Bill No. 2.338/2023 (acronym in portuguese “PL”), for the regulation of the use of Artificial Intelligence.

The Bill proposes the creation of the National Artificial Intelligence Regulation and Governance System (acronym in Portuguese “SIA”), which will include CADE, acronym in Portuguese for the Administrative Council for Economic Defense (article 40, item II, point “c”). Generally speaking, the SIA will be a “regulatory ecosystem”, coordinated by the Executive Branch and made up of agencies and sectoral regulatory and self-regulatory bodies for implementing and overseeing compliance with the Law on the use of Artificial Intelligence

Various contributions were submitted to the first proposal for a substitute, by authorities directly involved in the proposed regulation, such as ANATEL, ANPD and CADE, as well as academic bodies and industry associations (IBRAC, IDEC, Lawgorithm, ITI, ABIMED, CNPq, FEBRABAN and others).

CADE’s contributions mainly seek to include mechanisms for collaboration between the authorities that make up the SIA, guaranteeing greater legal certainty for joint action. Examples of cooperation cited by CADE include:

1. the collaboration of the other SIA authorities with CADE if they detect possible violations of the economic order;
2. CADE’s prerogative to request from the other SIA authorities technical subsidies for the analysis of mergers involving the use of artificial intelligence or for the instruction of administrative proceedings of facts jointly investigated by the SIA entities;
3. joint investigations between SIA regulators into high-risk AI systems, in order to avoid contradictory commands;
4. remote access to documents and training data on high-risk AI systems; and
5. principles and guidelines to guide the exchange of information between SIA authorities (i.e. confidentiality of sensitive data, integrity and security of accessed data, transparency, etc.).

Another important change suggested by CADE is the creation of a regulatory sandbox, which would be an experimental and controlled regulatory environment to test and monitor the competitive impact of algorithms, applying different rules to companies of different sizes and domains.

This would allow economic agents to adjust to the regulatory burden, with incentives for competition and innovation in the use of artificial intelligence. This is because, in CADE’s view, there is a critical risk that the rise in regulatory costs will increase barriers to entry, preventing the development of new technologies, especially for SMEs and startups.

On this point, CADE considers that the Digital Markets Act (DMA), the European Union’s regulatory initiative, as well as other proposals for regulating digital platforms under discussion around the world, have opted for asymmetric regulatory enforcement approaches, concentrated on a few players who run large ecosystems – the gatekeepers (Meta, Microsoft, Amazon, Apple etc.). This is the regulatory model advocated by CADE, according to our last post on the subject.

CADE’s recommendations are in line with and complement the substitute initially suggested by Senator Eduardo Gomes, which provided in items VI and VII of article 41 for the (i) regulatory-sanctioning competence of the SIA for the development or implementation of the use of artificial intelligence systems for economic activities (which includes the classification of high-risk artificial intelligence systems, under the terms of articles 15 and 16), as well as (ii) supervisory attribution to the sandbox.

The vote on the Bill by the Temporary Committee on Artificial Intelligence (“CTIA”, in the portuguese acronym), initially scheduled for Wednesday of last week (12/06), was postponed after discussions in the Senate plenary (11). On Tuesday of this week (18), a new report was presented with a second proposal for a substitute by rapporteur Eduardo Gomes, accepting 27 of the 55 amendments presented up until the reading of the opinion, as well as contributions from nine other bills that were being processed together with the Bill.

The new report presented took on board CADE’s main contributions by providing for:

• joint investigations between sectoral authorities into high-risk systems that may violate provisions of the PL or sectoral legislation (article 48, paragraph 2 of the new report);
• the duty of bodies and entities that are part of the CIS to communicate to CADE information obtained during inspections that may be of interest for the application of Federal Law No. 12.529/2011 (art. 48, § 3);
• CADE’s power to order developers to access the training, validation and testing data used to develop high-risk AI systems (art. 48, § 4); and
• asymmetric regulation to encourage technological development by micro-enterprises, small businesses and start-ups (art. 66).

The vote on the Bill, on the other hand, will take time to resume: on Tuesday (18), the CTIA again postponed judgment on the bill, and approved a request presented by Senator Marcos Ponte (PL-SP) for five more public hearings to be held to instruct the discussion on various fronts: risk assessment, governance, self-regulation and good practices, inspection and sustainable innovation.

CGM continues to follow the new discussions on this project and its practical developments.

This bulletin is for information purposes only and should not be relied upon to obtain legal advice on any of the topics dealt with here. For additional information, please contact the leaders of the Antitrust, Administrative, and Anti-Corruption Law team.

CGM Advogados. All rights reserved.