Digital Governance: Artificial Intelligence as an Ally – and as a Risk

Danielle Santos Figuerôa; Larissa Paula Stachio

DOI: 10.22167/2675-6528-202602812

Article derived from a Final Course Work (TCC), with content based on the student’s original work and adapted to the editorial format of the E&S Magazine with the support of the ResumeAI tool, an artificial intelligence solution developed by the Pecege Institute for textual synthesis and organization.

Abstract

The rapid rise of Artificial Intelligence (AI) has imposed unprecedented ethical and regulatory challenges, requiring organizations to review their control models to ensure business sustainability. Given the regulatory gaps and risks associated with AI adoption, the study investigated how the application of this technology impacts digital governance from the perspective of the ESG (Environmental, Social and Governance) pillars. The methodology adopted a qualitative-quantitative and exploratory approach, consisting of a documentary analysis of fifteen global case studies, application of a structured questionnaire with 65 professionals, and a semi-structured interview with an expert from a global technology organization. The results highlighted a critical compliance gap, as 75% of respondents reported the absence of clear guidelines in their institutions, which corroborated the concern of 45% of the sample regarding the lack of AI regulation. This scenario contrasted with the perception of governance as an asset of trust and a competitive differentiator. The research identified that risk mitigation, such as algorithmic bias and data security, required the implementation of continuous human supervision and the adoption of new lifecycle frameworks. It was also demonstrated that the convergence between AI and ESG criteria enhanced gains such as energy efficiency and social transparency. It is concluded that AI governance, when integrated with ESG principles, constitutes a strategic element for promoting compliance, transparency, institutional trust, and organizational sustainability.

Keywords: Algorithmic Ethics; Risk Management; Digital Governance; Artificial Intelligence; Corporate Sustainability.

1. Introduction

Digital transformation has driven significant changes in corporate and institutional environments, fostering the growing adoption of emerging technologies, particularly Artificial Intelligence (AI). AI is understood, according to the Organisation for Economic Co-operation and Development (OECD, 2024), as a set of computational tools that process input data to generate outputs such as estimates, recommendations, or decisions capable of influencing real and digital contexts. The application of this technology, in the view of Fernandes, Costa, and Abreu (2025), redefines organizational strategies, ranging from the automation of operational tasks to the execution of complex predictive analyses.

However, this advanced technological integration does not occur without the need for structures that guarantee the integrity and ethical direction of automated decisions. In this context, governance is established, defined by Loureiro Coutinho (2024) as the implementation of normative structures and accountability mechanisms that consider not only technical aspects but also legal and social impacts. In the realm of technology, AI governance emerges as a global, collaborative, and adaptable effort, according to a report by the United Nations (UN, 2024), based on human rights and sustainability, aiming to address complex risks and promote shared benefits.

Despite the strategic potential of AI, Fernandes, Costa, and Abreu (2025) warn about the ethical and regulatory concerns associated with its implementation, especially regarding information security, data privacy, algorithmic bias, and automated decision-making. The absence of robust governance mechanisms can lead to operational failures, reputational impacts, and financial losses for organizations. This issue is amplified by the compliance gap and concerns about the lack of clear regulation, as evidenced in the approved abstract, which points to a critical compliance gap and the absence of clear guidelines in many institutions.

The need for ethical governance of artificial intelligence is reinforced by the discussion on the importance of explainability of automated decisions and the prevention of discriminatory biases (Pacheco et al., 2023). Loureiro Coutinho (2024) emphasizes the urgency of public and corporate policies focused on civil liability and social protection, promoting algorithmic transparency. Although AI offers significant gains in efficiency and innovation, Floridi et al. (2018) highlight that it also intensifies ethical dilemmas, power asymmetries, and systemic risks, challenging traditional models of control and accountability. The absence of contextualized governance can compromise data security and privacy, perpetuate inequalities, and weaken institutional trust.

To structure a response to these risks, the ESG (Environmental, Social and Governance) pillars present themselves as a normative framework capable of guiding technological responsibility. The “United Nations Global Compact” (2004) establishes that ESG offers criteria for integrating environmental, social, and corporate governance issues into risk management and value assessment of organizations. Thus, when applied to Artificial Intelligence, ESG functions as a control tool that seeks to balance innovation with algorithmic transparency, privacy, and social justice.

Given this scenario of rapid technological transformation and unprecedented ethical and regulatory challenges, it becomes imperative to investigate the application of Artificial Intelligence in a context of digital governance aligned with ESG principles. Therefore, this research aims to investigate how Artificial Intelligence impacts digital governance aligned with ESG (“Environmental, Social and Governance”) principles, analyzing the potential of AI as a strategic tool and the risks associated with its adoption, focusing on ethical and social responsibility (privacy and bias) and on strengthening governance mechanisms (security and compliance) in automated decision-making.

2. Material and Methods

The present study adopted a qualitative-quantitative and exploratory methodological approach, characterized as a multiple case study complemented by bibliographic and documentary research. The objective was to investigate the application of Artificial Intelligence in a digital governance context aligned with ESG principles, analyzing its strategic potential and the risks associated with its adoption.

Secondary data collection was carried out through documentary analysis of public and internal materials. For this purpose, fifteen global case studies were selected, based on the relevance of the compliance incidents reported between 2022 and 2025. Of these, seven cases represented failures in AI governance and eight exemplified positive impacts of the technology.

The documentary analysis covered several sectors and geographies, focusing on four main areas of failure: algorithmic bias, data privacy, transparency, and accountability. The incidents were categorized by the nature of the risk and their ethical-legal impact, without nominal identification of the organizations, to support the discussion on the absence of compliance structures.

The collection of quantitative primary data was carried out using a structured, cross-sectional, and bilingual (Portuguese and English) questionnaire, containing 46 questions (23 per language). The instrument was divided into three blocks: “General Sample Overview” (14 questions), “Technical Perspective: System Creators and Maintainers” (5 questions), and “Governance Perspective: Users and Supervisors” (4 questions), using multiple-choice and Likert scale formats.

The questionnaire sample consisted of 65 professionals, with 61 interactions in Portuguese and 4 in English, recruited by convenience from professional networks and technology and business management groups. Data collection occurred between December 15 and 20, 2025, anonymously, via the Google Forms platform, ensuring data impartiality and privacy. The possibility of selection bias due to non-probabilistic sampling is acknowledged.

As a complementary qualitative step, a semi-structured interview was conducted with an expert in AI Automation and Data Platforms from a global technology organization (Appendix 2). The interview sought a strategic vision on AI governance, focusing on the technology lifecycle, from design to discontinuation, and on the perception of governance as a competitive differentiator.

The data analysis combined descriptive techniques for the questionnaire and content analysis for the interview. Additionally, an exploratory documentary research was conducted, which included annual sustainability and ESG reports (2023-2025), decisions from international regulatory bodies, journalistic reports, and legislation on AI. This procedure allowed for the confrontation of respondents’ perceptions with market practices and evidence of ethical failures.

The intersection of primary data, referring to participants’ perceptions, with secondary data, relating to market facts and cases, enabled a critical analysis of the effectiveness of the applied governance models. Ethical considerations were rigorously observed, with the guarantee of anonymity and confidentiality for all participants, in accordance with research principles.

3. Results and Discussion

The analysis of the collected data revealed a complex and multifaceted scenario regarding the governance of Artificial Intelligence (AI) in organizations, highlighting both the challenges inherent in its rapid evolution and the strategic potential of an approach aligned with ESG principles. The results evidenced a critical compliance gap and the urgent need for clear guidelines, contrasting with the perception of governance as a trust asset and a competitive differentiator. The research explored the perceptions of professionals, practical cases of failures and successes, and the global regulatory landscape, as detailed in the following analysis blocks.

Initially, the questionnaire applied to 65 professionals revealed that data privacy, with 26% of mentions, and algorithmic bias, with 28%, are the most critical concerns in the adoption of AI from a compliance perspective. Additionally, 45% of respondents identified the lack of clear regulation as the biggest challenge in aligning AI with organizational compliance policies. These findings underscore the urgency of addressing the ethical and legal risks associated with the technology, as also pointed out by Fernandes, Costa, and Abreu (2025), who emphasize the need for robust governance structures to mitigate operational failures and reputational damage.

The general perception of the sample indicated that 71% of participants believe that the benefits of AI outweigh the risks, and 82% associate AI integration with a significant increase in productivity and efficiency. However, this optimism is tempered by uncertainty about responsibility for AI-generated decisions, with 58% of respondents stating they have already reflected on the topic, but without clarity. Trust in AI recommendations also proved to be moderate, with 60% of participants assigning a neutral score (Score 3), while 34% expressed high trust (Score 4) and 6% low trust (Score 2).

The explainability of AI systems has emerged as a crucial factor for trust, with 80% of respondents agreeing that greater explainability increases confidence in the technology. However, the survey revealed that 75% of participants do not know where to report biased results, and 65% stated that their organizations do not offer training on the ethical use and compliance of AI. This gap in training and reporting channels weakens institutions’ ability to proactively manage risks, such as algorithmic bias, which, according to Silva (2022), is not an isolated technical error but a reproduction of structural inequalities.

From a technical perspective, aimed at creators and maintainers of AI systems (n=9), 55% of professionals did not participate in the development or integration of AI solutions, while 45% did. The biggest technical challenges to ensure compliance were identified as regulatory compliance (36%) and algorithm transparency (36%), followed by model explainability (28%). The majority of these professionals (72%) are unaware of or do not have access to tools that assist in AI explainability, which hinders the implementation of effective governance practices and the mitigation of complex risks, such as discriminatory biases.

Still from a technical perspective, the reception of clear guidelines on compliance requirements in AI projects is inconsistent: 36% of respondents receive them “sometimes” and another 36% “rarely”, with only 10% receiving them “always”. This variability in the communication of compliance requirements can compromise the integrity of systems from the initial development phases. The perception of the impact of internal audits on the AI development cycle was mostly neutral (53% Rating 3), indicating that, although they exist, audits may not be perceived as fully effective or integrated into the process.

The governance perspective, focused on users and supervisors (n=56), revealed that 63% of respondents have already made decisions based on AI recommendations. However, 76% have refused to follow an AI recommendation due to lack of trust, evidencing a latent distrust that can impact the adoption and effectiveness of the technology. The main compliance concerns for this group were privacy (48%), risk of error (35%), and lack of transparency (17%), reinforcing the need for mechanisms that ensure the integrity and explainability of the systems.

The understanding of compliance policies applied to AI systems used by users and supervisors also showed variation, with 39% assigning a neutral score (Score 3) and 22% a very low score (Score 1). This dispersion suggests that existing policies may not be clear, accessible, or sufficiently communicated, which hinders adherence and accountability. The absence of a uniform understanding of compliance guidelines can lead to diverse interpretations and, consequently, inconsistent practices in AI usage, increasing the risks of non-compliance and ethical failures.

The documentary analysis of fifteen global case studies complemented the respondents’ perceptions, showing that the absence of a compliance framework can result in severe damages. Cases of failure included gender bias in recruitment systems of an American multinational, civil liability for chatbot “hallucinations” in a Canadian aviation company, and the practice of “AI washing” by a surveillance technology company. These incidents demonstrate the materialization of risks of algorithmic bias and decision-making failures, which were pointed out as critical by 28% of respondents in the field survey.

In Brazil, reality presented parallels with global risks, such as automatic rejections at INSS, which generated a substantial increase in administrative appeals, and cases of algorithmic racism in public security systems, resulting in erroneous identifications and undue arrests. These episodes reinforce the thesis that algorithmic explainability, considered essential by 28% of technical participants, is not just a technical requirement, but a safeguard against automated discrimination that violates the principles of digital governance and the social license to operate for companies.

In contrast, the research identified eight success cases where AI was implemented with robust governance, generating value and efficiency. Examples include an American multinational that developed its own internal “Ethical Constitution,” a Brazilian company that used AI to streamline compliance investigations without losing ethical rigor, and a Brazilian bank that utilizes ethics committees to ensure transparent and bias-free credit models. These cases demonstrate that governance, when well-structured, transforms AI into a strategic ally, mitigating risks and promoting institutional trust.

The systematization of identified risks, according to the typology of Artificial Intelligence applied, revealed that digital governance cannot be treated generically. Predictive AI, used in recruitment or credit granting systems, requires a rigorous focus on the equity of historical data to meet the Social pillar of ESG, preventing algorithmic bias and structural discrimination. The interviewed specialist reinforced that the guarantee of fair decisions requires continuous human supervision, and not just the improvement of technical models, consolidating the social dimension as fundamental for maintaining the social license to operate for companies.

On the other hand, Generative AI, exemplified by chatbots and language models, demands transparency and traceability mechanisms that support the Governance pillar, especially to mitigate risks of data “hallucination” and privacy violation. Agentic AI, which transcends mere content generation to plan and execute tasks autonomously, imposes a new challenge to compliance: the governance of autonomy, where the traceability of actions becomes as critical as data accuracy, requiring compliance to be native in each agent developed.

The dimension of the Environmental Pillar demonstrated how AI can be a sustainability tool, mitigating its own energy cost. Google, for example, used DeepMind algorithms to optimize the cooling of its data centers, reducing energy consumption by 40%. Other cases include BMW, which uses AI and computer vision for recycling vehicle components, and JBS, which monitors suppliers via satellite to ensure the socio-environmental traceability of its production chain, addressing risks of deforestation and slave labor. These practical examples illustrate the potential of AI to achieve global decarbonization and natural resource optimization goals.

From the perspective of the Governance Pillar, respondents identified compliance practices as facilitators of technological adoption, reinforcing Mittelstadt’s (2019) argument that ethical principles, by themselves, are insufficient without concrete mechanisms for governance, monitoring, and accountability. The interviewed expert highlighted that governance acts as a “trust asset”, allowing the organization to stand out for the security and integrity of its solutions in a market saturated with new “players” without validated controls.

The absence of clear guidelines or specific training in AI, reported by 75% of users, weakens strategic governance. IBM’s 2025 Cost of Data Breach Report (2025) corroborates this gap, indicating that 63% of companies lack an AI governance policy. This lack of training points to the neglect of education and corporate culture, fundamental pillars for effective governance, where a workforce educated on the daily use of technology is essential for compliance policies not to be ignored at the operational level.

One of the most relevant findings of the research, present in both the documentary analysis and the questionnaire, is the negative impact of the lack of clear regulation. This factor was pointed out by 45% of respondents as the biggest challenge for aligning AI with organizational compliance policies. This perception corroborates the thesis of Jobin, Ienca, and Vayena (2019), who highlight how the absence of binding global regulatory frameworks creates a scenario of legal uncertainty, making it difficult to implement operationalizable ethical and governance guidelines in companies.

The absence of standardized norms makes it difficult to assign responsibilities, generating uncertainty in 58% of respondents about who should be accountable for failed automated decisions. In Brazil, Bill No. 2338/2023 (Brazil, 2023) emerges as a path to resolution, acting in conjunction with the LGPD to oversee data processing by artificial intelligence. By classifying recruitment and credit assessment systems as “High Risk”, the Brazilian Bill directly addresses the fears of bias and decision failures identified in the sample, requiring the implementation of Algorithmic Impact Assessments.

The global regulatory landscape presents three main models that influence Brazilian legislation. The EU AI Act (European Union, 2024) is the world’s first comprehensive law, adopting a risk-based approach and requiring full transparency for foundation models, serving as the main inspiration for the Brazilian Bill. The United States follows a sectoral and national security-based approach, focusing on consumer protection and the mitigation of catastrophic risks, without a single federal law (FTC, 2025). China, in turn, focuses on sovereignty and content control, regulating recommendation algorithms and generative AI and requiring systems to reflect social values (Kaufman and Zavaglia, 2024).

This global convergence towards regulation validates the findings, indicating that transparency and compliance are not merely ethical desires but essential regulatory requirements for companies in the digital age. The consulted expert projects a scenario of significantly stricter controls on data privacy and platform technical capabilities in the coming years. Thus, alignment with frameworks such as the EU AI Act or the future Brazilian Bill 2338 becomes the basis for the social license to operate, ensuring that technological innovation occurs within clear and safe legal boundaries for society.

In summary, the research demonstrated that AI governance is a strategic element for organizations, capable of transforming ethical and regulatory challenges into opportunities for strengthening trust and sustainability. The integration of ESG principles into AI governance not only mitigates risks such as algorithmic bias and privacy violation but also enhances gains in energy efficiency and social transparency. Digital compliance and social responsibility emerge as essential requirements for the social license to operate and the generation of ethical value in automated decision-making, demanding continuous human supervision and the adoption of new lifecycle frameworks.

4. Conclusion

The study investigated how Artificial Intelligence impacts digital governance aligned with ESG principles, analyzing its strategic potential and associated adoption risks, with a focus on ethical and social responsibility, and the strengthening of governance mechanisms. A critical compliance gap was found in organizations, where most professionals reported the absence of clear guidelines and expressed concern about the lack of AI regulation. It was identified that risks such as algorithmic bias and data security are central concerns, requiring continuous human oversight and the adoption of new lifecycle frameworks for mitigation. It was observed that governance is perceived as a trust asset and a competitive differentiator, and that the integration of AI with ESG criteria enhances gains in energy efficiency and social transparency. The main contribution lies in demonstrating that AI governance, when aligned with ESG principles, constitutes a strategic element for promoting compliance, transparency, institutional trust, and organizational sustainability, ensuring the social license to operate and the generation of ethical value in automated decision-making.

Despite the findings, the study presented limitations stemming from the use of convenience sampling and the predominance of descriptive analyses, which restricts the generalizability of the results to other populations. The sample composition and the temporal scope adopted also reflect a specific context of perception regarding AI governance, recommending caution in extrapolating to other organizational environments. For future research, it is suggested to expand the sample and use inferential statistical techniques, as well as conduct sectoral or comparative studies between countries. Such approaches could deepen the understanding of the complex relationships between AI governance, regulatory compliance, and ESG criteria, further consolidating the understanding of how technological innovation can be managed ethically and responsibly.

Bibliographic References

Brasil. 2023. Projeto de Lei nº 2338, de 2023. Dispõe sobre o uso da Inteligência Artificial. Senado Federal, Brasília, DF, Brasil. Disponível em: https://www25.senado.leg.br/web/atividade/materias/-/materia/157448. Acesso em: 28 jan. 2026.

European Union. 2024. Regulation (EU) 2024/1689 (Artificial Intelligence Act). Official Journal of the European Union, Bruxelas, Bélgica. Disponível em: https://eur-lex.europa.eu/legal-content/PT/TXT/?uri=CELEX:32024R1689. Acesso em: 28 jan. 2026.

Federal Trade Commission [FTC]. 2025. FTC Finalizes Order with IntelliVision over Al Washing Claims. Washington, DC, EUA. Disponível em: https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-finalizes-order-intellivision-over-ai-washing-claims. Acesso em: 12 nov. 2025.

Fernandes, A.A.; Costa, I.; Abreu, V.F. 2025. Governança da Inteligência Artificial: Estrutura, desafios e práticas para organizações inovadoras. Brasport, Rio de Janeiro, RJ, Brasil.

Floridi, L. et al. 2018. Al4People: An Ethical Framework for a Good Al Society: Opportunities, Risks, Principles, and Recommendations. Minds and Machines. Disponível em: https://doi.org/10.1007/s11023-018-9482-5 Acesso em: 10 ago. 2026.

International Business Machines Corporation [IBM]. 2025. Cost of a Data Breach Report 2025. IBM Security, Armonk, NY, EUA. Disponível em: https://www.ibm.com/reports/data-breach. Acesso em: 28 jan. 2026.

Jobin, A., lenca, M., Vayena, E. 2019. The global landscape of ai ethics guidelines. Nature Machine Intelligence. Disponível em: https://doi.org/10.1038/s42256-019-0088-2. Acesso em: 29 jan. 2026.

Kaufman, D.; Zavaglia, A. 2024. Governança e Ética de Inteligência Artificial nas Companhias Abertas. ABRASCA, São Paulo, SP, Brasil. Disponível em: https://s3-sa-east-1.amazonaws.com/abrasca/GovernancaEtica/8862-Relatorio-Final-Governanca-e-Etica-da-IA-nas-Cias-Abertas-Abrasca-Kaufman-e-Zavaglia.pdf. Acesso em: 16 jan. 2026.

Loureiro Coutinho, 2024 [Referência completa não encontrada no documento original]

MITTELSTADT, B.D. 2019. Principles Alone Cannot Guarantee Ethical Al. Nature Machine Intelligence, 1(11):501-507. Disponível em: https://doi.org/10.1038/s42256-019-0114-4 Acesso em: 08 ago. 2026.

OCDE, 2024 [Referência completa não encontrada no documento original]

Organização das Nações Unidas [ONU], 2024. Governing Al for humanity: final report. Disponível em: https://www.un.org/sites/un2.un.org/files/governing_ai_for_humanity_final_report_en.pdf Acesso em: 22 out. 2025.

Pacheco et al., 2023 [Referência completa não encontrada no documento original]

Silva, T. 2022. Racismo Algorítmico: Inteligência Artificial e Discriminação nas Redes Digitais. Edições Sesc, São Paulo, SP, Brasil.

UNITED NATIONS GLOBAL COMPACT. 2004. Who cares wins: connecting financial markets to a changing world. Geneva: UN Global Compact. Disponível em: https://documents1.worldbank.org/curated/en/280911488968799581/pdf/113237-WP-WhoCaresWins-2004.pdf. Acesso em: 15 mar. 2026.

Article originating from the Final Course Work of the Specialization in Compliance and ESG of the MBA USP/Esalq

To learn more about the course, click here and access the MBX Academy platform