Governança corporativa como elemento estruturante da gestão de riscos e conformidade legal e ética da inteligência artificial

• Ano de defesa: 2025
• Autor(a) principal: Coelho, Alexandre Zavaglia Pereira
• Orientador(a): Kaufman, Dora
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica de São Paulo
• Programa de Pós-Graduação: Programa de Pós-Graduação em Tecnologias da Inteligência e Design Digital
• Departamento: Faculdade de Ciências Exatas e Tecnologia
• País: Brasil
• Palavras-chave em Português: Inteligência artificial; Governança; Gestão de riscos; Ética; Regulação; Conformidade; Compliance; Explicabilidade; Justiça; Vieses; Framework
• Palavras-chave em Inglês: Artificial intelligence; Governance; Risk management; Ethics; Regulation; Compliance; Explainability; Fairness; Bias; Framework
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS
• Link de acesso: https://repositorio.pucsp.br/jspui/handle/handle/45371

The expanding adoption of Artificial Intelligence (AI) has emerged as a vector of economic and social transformation, owing to its capacity to classify and correlate vast datasets, thereby supporting more precise and strategic decision-making processes, and enabling the generation of synthetic content—including texts, videos, and audio—which constitutes both a significant innovation for optimizing professional activities and enhancing productivity, and an instrumental resource for developing novel products, services, and business models. Beyond AI development companies, Application Sectors (ASs) have demonstrated a pivotal role in the expansion and adoption of this technology, exercising their innovative complementarity. Within this context, given its sociotechnical nature, AI presents a duality between benefits and risks concerning established rights, health, and safety of individuals, groups, and society at large. Its characteristics and limitations, particularly through the predominant utilization of deep neural network architectures and unsupervised training methods, among other specificities, have directly impacted issues pertaining to transparency, explainability, control, security, accountability, justice, and equity, thereby propelling agendas to establish general ethical principles (Ethics for AI), transform them into operational safeguards (Responsible AI), and develop movements aimed at addressing technical issues of safety (AI Safety), explainability (xAI), and fairness (Fairness in AI). Nevertheless, tensions between commercial and social interests, coupled with the inadequacy of self-regulation strategies, have necessitated state regulation, exemplified by the EU AI Act, which incorporated ethical principles and established foundations for risk management and corresponding governance obligations for each risk level, including mechanisms for monitoring, oversight, auditing, and imposing sanctions for noncompliance. Despite these advances, numerous organizational challenges persist in translating general and abstract ethical principles, alongside regulatory guidelines that reflect AI's inherent complexity, into operational practices and safeguards. In this scenario, the importance of organizations in achieving regulatory effectiveness is reinforced, as is corporate governance as an orchestrating force to ensure responsible and sustainable AI use. However, various questions remain regarding how to navigate this complexity, what essential themes and elements must be observed for structuring corporate AI governance areas, their structure—whether through organic structures or specialized subsets—their formation, frameworks, policies, and protocols, among other aspects. In this context, this publication, directed toward Application Sectors (ASs) that predominantly act as AI implementers, seeks to analyze the nuances that technology professionals must observe regarding sociotechnical aspects and, conversely, the AI characteristics and limitations that others must observe to translate abstract concepts and definitions into objective decisions and criteria, helping to think through and solve practical, real-world problems. An approach that designates corporate governance as an orchestrating function, as a structuring element for AI risk management and ethical and legal compliance activities