Kakao has established ethical principles and implementation guidelines for AI services, applicable from planning to operation, in response to the rise of autonomous agents. The company aims to minimize risks associated with AI's ability to plan and act independently by managing human control, accountability, and the authority granted to AI, while also creating a system to inspect risk factors throughout the service lifecycle.
On October 13, Kakao unveiled its 'AI Ethics Principles and Implementation Guide for the Era of Autonomous Agents' during the developer conference 'IfKakao·if(kakao)26' held at the Kakao AI Campus.
Hajin Hwa, Kakao's Chief of AI Communications, emphasized the need to define human judgment and responsibility in environments where AI operates autonomously, and to incorporate ethics into service design and operations.
The new AI ethics principles, termed 'SPIRIT,' consist of six components: Safety, Privacy, Inclusion, Responsibility, User Agency, and Transparency and Trust. These principles adopt the three core values and seven key principles from the 'Korean AI Ethics Principles' as primary implementation standards, which have been further tailored to fit Kakao's services.
Kakao has developed a framework called 'NORMA' to implement these ethical principles in actual services. This framework aims to manage complex AI ethical challenges at the organizational level and to translate them into actionable steps.
The ethical principles were developed with input from a seven-member external advisory group of the AI & Technology Ethics Subcommittee, which was established in July. The advisory group ensured alignment with domestic and international norms, while Kakao's departments related to AI safety, privacy, information security, and services reviewed the principles from a practical perspective.
Kakao categorizes the risks associated with autonomous agents into five management areas: Control Authority, Objectives, Permissions, Reference Resources, and Interactions, linking necessary controls such as human oversight, emergency control, authority management, and data protection.
Based on this framework, an 'AI Ethics Implementation Checklist' will operate in three stages: identifying service attributes, pre-risk screening, and checking and assessing control items. This checklist is expected to be implemented by the end of the year.
Kakao will also enhance its measures to protect children and adolescents. In addition to the 'Digital Child and Youth Protection Checklist' implemented since 2025, the company plans to incorporate AI-specific child risk assessment elements based on the OECD and UNICEF's online risk classification system (5Cs) into the new checklist.
Kim Se-woong, chair of Kakao's AI & Technology Ethics Subcommittee, stated, “We have established a risk review system that operates throughout the entire lifecycle of AI services, incorporating perspectives from both the external advisory group and internal organizations. As AI autonomy increases, we will clarify human judgment and responsibility, and create AI services that users can trust and feel safe using.”
Meanwhile, last month, Kakao signed a memorandum of understanding with the AI Safety Research Institute to jointly evaluate AI safety and establish an evaluation system. The two parties will collaborate on assessing the safety of AI models and agents and developing evaluation tools to identify risk factors.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.