Challenges to the regulation of artificial intelligence in Brazil
Bruna Borghi Tomé
TozziniFreire Advogados, São Paulo
Carla do Couto Hellu Battilana
TozziniFreire Advogados, São Paulo
Sofia Gavião Kilmar
TozziniFreire Advogados, São Paulo
Artificial intelligence (AI) has been gaining traction in many sectors and industries since 2000. Its business applications have expanded considerably across several countries, including in Brazil. The rapid growth in AI applications has sparked discussions around how this technology should be regulated.
Today, there are myriad circumstances in which AI can be used. It can be fully integrated in, for example, virtual assistants, streaming platforms, investment robots, self-driving cars, social networks, biometrics and even in medical diagnosis software. The rapid growth in its use for multiple systems, as result of increasing techno-scientific evolution, had generated a vast array of challenges for the law when it comes to keeping up with its response in an ever-changing AI reality.
Futuristic problems concerning the regulation of AI are now today’s problems. To what extent do existing norms and laws affect AI and are they enough to regulate it? If new specific regulation comes into force, what would be its effect on businesses already using AI systems? Would such effect be beneficial or harmful? Would this new regulation, depending on how it is integrated into the legal system, meet or go beyond the desired protection? While such questions have become increasingly prominent in the Brazilian legal environment, AI application in products and services available to the Brazilian public in the meantime continue to grow in scale and complexity.
ln this context, in 2021 the Bill of Law No 21 of 2020 (‘Bill No 21’) was presented before the Brazilian National Congress. The Bill was proposed as the legal framework for regulating AI in Brazil, placing the country in a select group of states with concrete AI regulation laws: China, members of the European Union, Singapore and the United States. Following approval of the revised version by the Brazilian National Chamber of Deputies on 29 September 2021, Bill No 21 was sent to the Brazilian National Senate and is expected to be approved later this year.
The Bill’s main inspirations are (i) the Recommendation on Artificial Intelligence by the Organization for Economic Co-operation and Development (the ‘OECD Recommendation’); and (ii) the discussions held in the EU, which inaugurated the world's first proposed Artificial Intelligence Act (the ‘EU AI Act’).
Bill No 21 took inspiration from the OECD Recommendation by establishing guiding principles, such as transparency, human wellness as the main goal of AI, right to privacy, and user security as a priority, among others.
The EU AI Act takes a risk-based approach, regardless of the specific market destination of the AI system in question, in order to protect users from unacceptable use of AI. The proposed Act establishes a solid methodology for analysing and classifying risks, determining criteria for identifying the so-called ‘prohibited risks’ and ‘high risks’ for the use of AI (Titles II and III, respectively).
Furthermore, the EU AI Act was designed to fit harmoniously within the sectoral policies already in effect across the EU. In the words of the legal document itself, such a strategy aims to ‘ensure consistency, avoid duplication and minimise additional burdens’ so that AI regulation ensures dialogue and compatibility, for instance, with EU competition and product safety regulations.
At a 2021 public hearing held at the Brazilian National Chamber of Deputies, Bill No 21 was also raised to bring attention to how a greater degree of detail in specific sectorial regulations may yet be achieved. As detailed regulation and suitable sanctions for infringement may change depending on the sector (for instance, in the banking industry or consumer goods), Bill No 21 allows for detailed provisions to be established by sectorial authorities, considering the specific aspects of each sector and AI’s applicability within it.
The goal of Bill No 21 is to encourage AI research and foster AI innovation within the public and private sectors, and at the same time protect the population, especially with regard to privacy and transparency. Given this ambitious goal, the Bill shall be subject to further debate at the Brazilian National Senate among panels of experts representing both the public and private sectors.
Although discussions will likely extend over the next few months, one thing is certain: a regulatory framework for AI in Brazil is just a matter of time. It is imperative that legislative bodies fully grasp the different aspects and nuances of complex systems like AI in order to regulate them correctly.