0
Kakao unveiled its AI ethics principles and implementation guidelines for the era of autonomous agents. / Kakao

Kakao has established new AI ethics principles and an implementation framework to be applied from service planning through operation, aligned with the proliferation of AI agents that autonomously plan and act. As AI autonomy grows, the company plans to build a system that verifies human oversight, accountability, and authority controls across the entire service lifecycle.

According to Kakao on the 15th, the company unveiled its new AI ethics principles, "SPIRIT," and operational framework, "NORMA," at the developer conference "if(kakao)26" held at the Kakao AI Campus on the 13th.

SPIRIT comprises six principles: Safety, Privacy, Inclusivity, Responsibility, Individual Autonomy, and Transparency & Trustworthiness. The framework adopted the South Korean government's "National Artificial Intelligence Ethics Principles" as its primary standard while tailoring detailed provisions to Kakao's service characteristics.

The operational method for applying these principles to actual services was established as "NORMA" (Navigating, Organizational, Responsibility, Manageable, Actionable). It focuses on managing complex AI ethical challenges based on shared organizational responsibility and translating them into actionable, executable steps.

The new ethical framework was finalized earlier this month in conjunction with seven external advisory panel members of the AI & Tech Ethics Subcommittee launched in July. The external advisors evaluated alignment with domestic and global AI norms, while internal teams handling AI safety, privacy, information security, and service operations reviewed practical feasibility.

In particular, risks that may arise from autonomous agents are categorized and managed across five domains: control initiative, objectives, authority, reference resources, and interactions. Each domain is linked to countermeasures such as human oversight, emergency override, authority controls, and data protection.

Operating on this foundation, the "AI Ethics Implementation Checklist" follows a three-stage process that first identifies service attributes, screens potential risks in advance, and then evaluates and decides on necessary control measures. Kakao plans to implement the new checklist across actual services within this year.

Oversight of AI services targeting children and adolescents will also be strengthened. Building on its existing "Digital Child and Youth Protection Checklist," Kakao is incorporating AI-specific risk factors into the new checklist, drawing reference from the "5Cs," an online child risk classification framework established by the OECD and UNICEF.

Since releasing its Algorithm Ethics Charter in 2018, Kakao formed the AI & Tech Ethics Subcommittee in 2022 with participation from Kakao and its key affiliates. It has continually expanded related frameworks, introducing group-wide guidelines including the "Guidelines for Responsible AI" and the "Key Checklist for Safe AI."

Kim Se-woong, chair of Kakao's AI & Tech Ethics Subcommittee, said, "By incorporating perspectives from both external advisory experts and internal organizations, we have established a risk review framework that operates across the entire AI service lifecycle." He added, "As AI autonomy expands, we will clearly delineate human judgment and responsibility to build AI services that users can trust and use with confidence."

                                                                                                          Son Kang-hoon