91Â鶹¹ú²ú¾«Æ·×ÔÅÄ

From school assignment algorithms in France to AI-assisted emergency response systems in Spain, automated administrative processes in Latin America, and AI-driven law enforcement in the UK—artificial intelligence (AI) is increasingly embedded in public sector decision-making. While these technologies promise efficiency and improved services, they also raise critical ethical concerns regarding transparency, bias, fairness, privacy, and accountability. 

In response to these challenges, UNESCO developed the Recommendation on the Ethics of Artificial Intelligence, adopted in 2021 as the first global framework guiding ethical AI development and implementation. One of its key tools is the Ethical Impact Assessment (EIA)—a structured framework designed to help institutions identify and mitigate ethical risks in AI systems. 

The EIA is a practical tool that enables governments, public institutions, and other organizations to assess the ethical implications of AI projects before and during their development and deployment. By promoting proactive governance, it helps ensure that AI systems align with ethical values and fundamental rights. The assessment covers: 

Core Ethical Values and Principles 

The EIA framework is structured around scoping project questions and key ethical principles outlined in UNESCO’s Recommendation, including: 

• Safety and Security 

• Fairness, Non-Discrimination, Diversity 

• Sustainability 

• Privacy and Data Protection 

• Human Oversight and Determination 

• Transparency and Explainability; Accountability and Responsibility 

• Awareness and Literacy 

A Dynamic and Evolving Tool 

Given the rapid evolution of AI and its increasing complexity, the EIA is designed to be a living tool—one that undergoes continuous updates and refinements based on practical experiences and emerging ethical challenges. To enhance its effectiveness, the framework requires ongoing testing and feedback from real-world implementations, ensuring that it remains relevant and adaptable to diverse institutional needs. 

To refine the EIA and tailor it for practical use, UNESCO and GobLab UAI piloted the tool across Latin America, working with public and private institutions to gather critical feedback. Piloting participants included Bogotá City Hall, which applied the EIA to Chatico, an AI-based citizen service platform, analyzing its impact on privacy, transparency, and governance; Excelsis, a Paraguayan private tech consulting firm associated with CISOFT, which tested the EIA on a generative AI chatbot designed for corporate data management, identifying challenges related to interoperability, regulatory alignment, and business sector adoption; finally, five Peruvian institutions, coordinated by the Secretariat of Government and Digital Transformation, assessed AI systems in areas such as education, environmental certification, agriculture, and energy regulation, emphasizing the need for contextual adaptation, modularity, and institutional integration. 

The pilot projects provided valuable insights into how UNESCO’s Ethical Impact Assessment (EIA) can continue evolving to support institutions in better fostering responsible AI development, and these findings are already being used to revise and improve the tool. The application of the EIA proved to be an eye-opening and deeply engaging experience for participating teams. Many described the process as transformative, providing a structured opportunity to reflect critically on the ethical dimensions of AI that had previously been overlooked. 

Participants expressed enthusiasm for learning more and strengthening their institutional approaches to AI governance. Beyond compliance, the EIA fostered a genuine commitment to responsible AI, with teams feeling empowered to champion ethical innovation within their organizations. This sense of ownership and motivation highlights the EIA’s potential not only as an assessment tool but also as a catalyst for ethical leadership in AI adoption worldwide.  

According to participants, the tool’s potential is not only as an ethical evaluation framework but also as a catalyst for meaningful conversations and institutional commitment to AI governance. The process empowered teams to take ownership of ethical AI implementation, reinforcing the importance of fairness, transparency, and accountability at every stage of development. 

To ensure that institutions of all sizes and resource levels can fully leverage the EIA, participants recommended transitioning to a digital, interactive platform. A modular design would allow organizations to customize the assessment based on their specific needs, streamlining the process while maintaining its depth. Additionally, a simplified version of the tool would provide an entry point for institutions with limited AI expertise, making ethical evaluation more accessible and actionable. 

Recognizing the diverse regulatory and operational landscapes in which AI is deployed, institutions emphasized the importance of refining the EIA to align with national governance frameworks. Clearer, non-technical language would enhance inclusivity, while the integration of environmental and social impact indicators would provide a more holistic evaluation of AI systems. These enhancements would enable a wider range of organizations to incorporate ethical assessments into their existing structures seamlessly. 

Participants saw an opportunity to embed the EIA more deeply into governance and compliance processes, ensuring its long-term impact. To support this, institutions recommended the development of tailored training programs and ongoing capacity-building initiatives. By fostering interdisciplinary collaboration and establishing structured feedback loops, organizations can continuously refine their AI ethics strategies, aligning them with evolving challenges and best practices. 

To reinforce accountability and long-term monitoring, institutions proposed incorporating quantitative indicators within the EIA. A structured risk matrix could help assess ethical considerations at different stages of AI implementation, while follow-up mechanisms would ensure continuous oversight and improvement. By integrating these elements, the EIA can serve not only as a diagnostic tool but also as an instrument for sustained ethical AI governance. 

A strong foundation in AI ethics literacy was recognized as essential for the EIA’s success. Participants highlighted the need for structured training, hands-on workshops, and accessible learning materials to empower teams with varying levels of expertise. Expanding outreach efforts would also help engage a broader audience, from public officials to private sector leaders and civil society organizations. By making ethical AI governance a shared responsibility, the EIA can play a crucial role in shaping a culture of responsible AI adoption. 

As AI continues to shape public governance and critical infrastructure, ensuring ethical oversight and accountability is more important than ever. The EIA provides a structured approach to aligning AI innovations with fundamental values, supporting institutions in navigating the balance between technological progress and human-centered responsibility. 

By continuously refining and expanding its reach, the EIA is poised to become a global standard for ethical AI governance. With ongoing collaboration, innovation, and institutional commitment, AI can be developed and deployed in ways that are fair, transparent, and beneficial for all.