*This is an AI-powered machine translation of the original text in Portuguese

** Image resource obtained from  Freepik.com

Our partner Bernardo Fico recently commented on the recommendations presented by UNESCO in the public consultation on the Assessment of Readiness for Ethical and Responsible Artificial Intelligence (AI) in Brazil, a central topic for our office's activities. The recommendations represent a significant step towards ensuring the inclusion of human rights throughout the lifecycle of artificial intelligence systems in our country. They focus on guiding principles and values of ethics that should be integrated into the development and deployment of such technologies.

The Recommendation primarily utilized the Readiness Assessment Methodology (RAM), a analysis system developed by the UNESCO secretariat. It covers legal-regulatory, socio-cultural, economic, scientific-educational, and technological and infrastructural dimensions. The RAM provides a report indicating the state of the art in each country for each parameter, aiming to assess the readiness of nations regarding the ethical development and use of AI technologies.

In his commentary, Bernardo emphasized the importance of diversity and inclusion considerations in the formulation of automated recommendations. He highlighted the potential perpetuation of biases through the automation of biased decisions, making it essential to incorporate monitoring and internal mitigation measures into the recommendations. He suggests the adoption of concrete mechanisms to encourage diverse workforce in the AI sector, in conjunction with public policies promoted by government bodies such as the Ministry of Labor and the Ministry of Science, Technology, and Innovation.

In the educational context, Bernardo points out that the adoption of AI resources can be a relevant tool in knowledge development, provided the associated risks are mitigated. He emphasizes the importance of human involvement in the educational process but acknowledges the potential of AI tools as pedagogical instruments. Regarding the protection of children and adolescents, he highlights the relevance of developing and adapting AI systems that take into account the specific characteristics of this audience to prevent risks associated with their inherent vulnerabilities. Additionally, he underscores the potential of AI in protecting minors in virtual environments by detecting inappropriate content and identifying users who use online services with fraudulent age statements.

Bernardo also initiates a discussion about the role that AI technologies can play in environmental preservation, such as global environmental monitoring aimed at combating deforestation and protecting biodiversity. He references various governmental and private initiatives that use AI systems to enhance environmental monitoring and management, as well as planning enforcement operations and mitigating identified environmental issues.

He suggests the inclusion and promotion of new forms of contracting with the Public Administration to effectively incentivize technological development and innovation through partnerships with the private sector. He points out current obstacles in the development of AI technologies due to the inflexibility of existing contracting models, advocating for measures that ensure closer collaboration between the private sector and the government, aiming for the joint development of new solutions.

Furthermore, Bernardo emphasizes the importance of ensuring the compatibility of AI systems with human values and ethics. This concern requires continuous effort throughout the existence of AI technologies, and the desired compatibility can be achieved through design that incorporates audits and alerts to allow human intervention in case of atypical behaviors, as well as through the ongoing training of professionals involved in the development of these systems.

Finally, Bernardo dedicates attention to the need for documentation of risks and mitigation strategies applied throughout the development and lifecycle of Artificial Intelligence. These records should be continuously updated. However, the lawyer understands that the disclosure of reports on such vulnerabilities should only occur in specific circumstances, such as for audits and monitoring, to prevent the exploitation of these vulnerabilities and the consequent exposure of users. Thus, accountability measures should be established depending on specific circumstances, based on the assessment of risks involved and the differentiation of measures accordingly.

The UNESCO recommendations presented in the public consultation are a significant step in the pursuit of ethical and responsible AI in Brazil, and the comments on the recommendations contribute to enriching the debate on how to achieve this goal.