Artificial Intelligence: 10 BASIC PRINCIPLES ABOUT THE AI REGULATORY FRAMEWORK IN BRAZIL
The Purpose of the decalogue launched by FecomercioSP aims at the debate on how to reconcile disruptive innovations with ethics and long-term economic and social growth.
With the democratization of artificial intelligence (AI) and the impact that this has been causing, the debate on its regularization is growing worldwide. And with several aspects: in the European Union, the proposal is for a denser framework, which works on what are the corporate governance criteria based on AI; in the United States, there are ethical and responsible guidelines, specifying what type of governance should be adopted; in Brazil, the 2021 bill, approved by the Chamber, will be discussed in the Senate and tends to follow the EU trend.
“We want to consider AI legislation, how to protect fundamental rights and warranties and, at the same time, not impede technological evolution. The standard needs to be flexible and adaptable. In the case of the consumer electronics segment, companies that develop AI outside Brazil may not want to do the same in the country. It is necessary to raise the level of maturity of the discussion”, says Rony Vainzof, consultant in data protection and author of the decalogue together with Andriei Gutierrez, president of the Digital Economy Council, both from FecomercioSP.
POINTS TO BUILD AN ALGORITHMIC FUTURE
1 – REGULATE THE USE AND NOT THE TECHNOLOGY – Just like the steam engine, electricity and the internet, artificial intelligence (AI) is a general-purpose technology, which is in full and constant development. Extreme caution is needed when regulating technologies so that the standard is sufficiently flexible and adaptable to its rapid changes, allowing experimentation, innovation, and continuous evolution of AI systems. Please note amidst the global debate over AI regulation, we have already been ushered into a new era, that of generative AI. Thus, it is necessary to adopt a principled approach, like the Civil Rights Framework for the Internet (MCI), and a less prescriptive one, like the General Data Protection Law (LGPD), which protects not a technology, but, very well, the right and fundamental guarantee of the protection of personal data. Different objects demand different regulations to be more efficient in guaranteeing rights and legal certainty for innovation.
2 – EVALUATE THE APPLICATION OF THE LEGISLATION IN FORCE AND THE COMPETENCE OF THE CONSTITUTED BODIES – As AI is a general-purpose technology, according to its use, certain existing legislation will apply, such as the LGPD, the Consumer Protection Code (CDC) , the Civil Code, the Positive Registration Law and the Civil Rights Framework for the Internet, in addition to the Federal Constitution itself and sectoral market regulations. That is, if there are already applicable laws protecting individual rights and guarantees, which have not even been stressed for AI, we must have an agile governance model, which does not harm investment opportunities and does not impede technological innovation, like countries as United Kingdom, Japan, Australia and Singapore.
3 – OBSERVE THE EXISTING RULES ON CIVIL RESPONSIBILITY – Brazil already has a broad, robust, and consolidated civil liability regime, which is laid down in the most diverse legislation, such as the Civil Code, the Consumer Defense Code and the LGPD. These standards regulate and protect individual and collective rights in the event of damages arising from AI systems, according to their use. A new exclusive civil liability regime for AI would be reckless, given the risk of discouraging the development of AI in Brazil and/or access to cutting-edge technology produced abroad.
4 – OBEY THE CURRENT REGIME OF ADMINISTRATIVE SANCTIONS – According to the use of AI, there are already administrative sanctions to be applied by the competent sectoral bodies, such as in the Consumer Protection Code, in the LGPD, in the Civil Rights Framework for the Internet, among other norms. That is, any new sanctions regime would create a reckless bis in idem. For example, despite the important debates on the existence of best practices for mitigating discriminatory biases by software or automated systems, it should be remembered that the LGPD has non-discrimination as a principle, and the ANPD already has regulatory and sanctioning powers to the promotion, encouragement and adoption of best practices, as well as the safeguarding of fundamental rights involving personal data.
5 – RISK-BASED, PRINCIPLEOLOGICAL AND CONTEXTUAL APPROACH – The key point of the regulation should focus on modeling staggered levels of risks and creating proportional safeguards for their mitigation, depending on the contextual application of the AI. It is pertinent that, at the legal level, the general risk parameters are outlined so that their definition, within the normative scope, takes place in a contextual manner, weighing the potential risk and the mitigation measures available at the time of regulation. The evolution of AI has also been accompanied by the evolution and emergence of technical and technological measures to mitigate risks. Therefore, the effectiveness of the AI Legal Framework to protect rights in balance with the stimulus to innovation depends on the understanding and observance of this primordial element. It is proposed, therefore, that the standard brings criteria for risk assessment, but that they serve as guidelines for its use by the various sectors of the economy. The logic is: risk should not be endorsed by legislation. The AI sector is responsible for assessing the risk as technology matures in Brazil and more precisely identifies the risks involved in each activity or application in its context. Whoever develops or applies AI may be charged, inspected, sanctioned, and held accountable in case of mistaken risk assessment and absence of the duty of care, through the institute of regulated self-regulation.
6 – SOFT LAW COMBINED WITH REGULATORY FLEXIBILITY – Prioritize general governance guidelines, such as guidelines for ethical and responsible use, leaving the factual analysis to be done on a case-by-case basis in accordance with the concrete risk and with the understanding of the sectoral regulatory bodies. The choice of regulatory strategy must be consistent with the behavior displayed by regulated agents and permanently adapted and optimized, that is, responsive regulation. The success of regulating an object in constant transformation depends on the combination of soft law and regulatory flexibility, so that regulators and regulated parties can advance together in risk mitigation measures and actions.
7 – PROMOTING REGULATED SELF-REGULATION – Reconhecendo a dinâmica de evolução constante da tecnologia, a inovação e a diversidade de contextos para cada setor de aplicação, apoiamos a adoção de soluções de corregulação ou “autorregulação regulada”, provendo segurança jurídica e proteção de direitos fundamentais. A norma traz parâmetros mínimos de governança no desenvolvimento, emprego e monitoramento dos sistemas de IA, capazes de balizar o dever de cuidado a ser tomado. Esses parâmetros mínimos orientarão instituições de autorregulação no desenvolvimento de códigos de conduta específicos para diferentes setores de atividade econômica e de possível emprego da tecnologia, podendo ser reconhecidos posteriormente por autoridades públicas setoriais competentes.
8 – EMPLOYING DECENTRALIZED REGULATION WITH REGULATORY INTEROPERABILITY – A risk-based, principled, and contextual regulatory model for AI differs from proposed models based on a specific regulatory agency for AI. We understand that, in addition to not being effective, the regulatory model based on a centralized agency can bring more bureaucracy without guaranteeing the necessary protection of fundamental rights. Furthermore, it would be fiscally irresponsible to propose the creation of yet another government agency without being fully sure of its need and effectiveness. Brazilian society has seen, for years, the attempt to create and structure the ANPD, still without the necessary resources for its important mission. On the other hand, we defend that there should be a centralized governmental instance for the promotion of regulatory interoperability between the different competent authorities: sharing of nomenclatures, standards, best practices, and harmonization of regulated self-regulation between different sectors. In addition, we also defend the creation of a consultative body with the participation of the private sector and society, multisectoral.
9 – ENCOURAGE RESPONSIBLE INNOVATION – The proposed system should provide mechanisms to benefit virtuous agents. That is, those who use AI systems in accordance with established ethical and regulatory precepts and act preventively to avoid any type of use of their technology in a biased, discriminatory and inappropriate manner. With a view to building an ecosystem that leverages initiatives and encourages competition, we suggest that the standard promote development programs and opportunities for small businesses and startups, including through the creation of controlled innovation environments, with sandboxes and regulatory hubs.
10 – ADOPT GLOBAL STANDARDIZATION – One of the key characteristics of the digital economy is that its global value chains are inherently dynamic. It is of great importance that Brazil has an active participation and voice in international forums for the discussion of best practices and global governance of AI, especially so that we have convergence in terms of standards and regulations. The competitiveness of Brazilian organizations, as well as their access to new global technologies, will increasingly depend on an active Brazilian participation in the construction of these global policies. We recommend participating in initiatives led by the Organization for Economic Cooperation and Development (OECD), as well as those linked to the Global Partnership on AI (GPAI), the G7 and the United Nations Educational, Scientific and Culture (UNESCO).
Source: Revista Eletrolar News – Edição #155
