|
The Ministry of Science and ICT (MSIT) and the TTA announced the results of a crosswalk analysis between Korea's " Guidebook for Development of Trustworthy AI " and the U.S. National Institute of Standards and Technology (NIST) "AI Risk Management Framework (AI RMF)" on December 24
The Ministry of Science and ICT (Minister: Yoo Sang-im, hereinafter referred to as “MSIT”) and the Telecommunications Technology Association (Chairman: Son Seung-hyun, hereinafter referred to as “TTA”) announced the completion of a crosswalk analysis* between Korea’s Guidebook for Development of Trustworthy AI (hereinafter referred to as the “AI Guidebook”) and the U.S. National Institute of Standards and Technology (hereinafter referred to as “NIST”) AI Risk Management Framework (hereinafter referred to as “AI RMF”).
* The crosswalk analysis was conducted by TTA (research organization for the ‘AI Guidebook’) and NIST (research organization for the ‘AI RMF’) from February to December 2024.
※ Detailed results of the crosswalk analysis can be found on the following websites: TTA(https://www.tta.or.kr/eng/index.do), U.S. NIST(https://airc.nist.gov/AI_RMF_Knowledge_Base/Crosswalks), Korean AI Safety Research Institute(https://www.aisi.re.kr/eng)
Since 2021, MSIT and TTA have been working to ensure AI Trustworthiness by developing and distributing the AI Guidebook. This effort has outlined 15 technical requirements (with 67 detailed verification items) to support domestic companies in voluntarily achieving AI trustworthiness during the entire lifecycle of AI systems. Additionally, based on the AI Guidebook, MSIT and TTA have established the group standard "Requirements for Enhancing the Trustworthiness of AI Systems" (12.2023) and launched the "Voluntary Certification of AI Trustworthiness" program (11.2023), with a total of seven certifications issued). These efforts aim to support the creation of a sustainable AI innovation ecosystem by continuously advancing policies to ensure AI Trustworthiness and safety.
< Overview of Korea's AI Guidebook (02. 2022) >
-
Provides detailed technical requirements for ensuring supporting AI Trustworthiness and mitigating/responding to risks during the AI system development process, structured across the five stages of the AI lifecycle* to support the voluntary efforts of the private sector in achieving AI trustworthiness.
*Planning and Design - Data Collection and Processing - AI Model Development - System Implementation - Operation and Monitoring
NIST, a U.S. government agency leading research and guideline development on AI Trustworthiness, released the voluntary AI RMF in January 2023. The AI RMF aims to support individuals and organizations in understanding, managing, and mitigating risks related to the design, development, deployment, and use of AI systems. It has played a key role in global discussions and advancements on approaches to achieving AI Trustworthiness across both private and public sectors. Furthermore, numerous organizations worldwide have adopted and utilized the AI RMF as a foundational framework for achieving AI trustworthiness.*
* Examples of crosswalk analyses with the U.S. AI RMF include International Standard Organization’s ISO-IEC 42001: AI Management System, Singapore's AI Verify, and Japan's AI Guidelines for Business.
< Overview of U.S.'s AI RMF (01. 2023) >
-
A guideline document developed by the National Institute of Standards and Technology (NIST) under the 2020 National AI Initiative Act. It assists organizations involved in designing, developing, deploying, or using AI systems to manage risks, thereby promoting the development and use of trustworthy and responsible AI systems.
This crosswalk analysis was conducted to evaluate the characteristics of Korea's AI Guidebook and the U.S. AI RMF, two influential frameworks in the field of AI Trustworthiness, and to verify the alignment of detailed items. The analysis ed that a substantial portion of the detailed items align closely and that both guidelines provide complementary approaches to enhancing AI Trustworthiness and mitigating related risks.
* While there are some differences in specificity across items, the majority of them are harmonized (63 out of 67 detailed verification items in Korea’s AI Guidebook align with the U.S. AI RMF).
Furthermore, the crosswalk analysis with the NIST AI RMF has ed the global-level sophistication and completeness of Korea’s technical and verification framework for AI Trustworthiness. Based on the results of this analysis, MSIT and TTA plan to strengthen technical cooperation with NIST and expand efforts to ensure harmonization with global technical standards and norms in the field of AI Trustworthiness. Additionally, the harmonized aspects with the NIST AI RMF will be reflected in the Voluntary Certification of AI Trustworthiness to enhance global alignment and promote exemplary practices for voluntary AI Trustworthiness assurance among domestic companies.
TTA Chairman Son Seung-hyun stated, “This crosswalk analysis has been a turning point in elevating Korea’s AI Trustworthiness framework to a global standard. Going forward, we will continue to expand global cooperation to enhance the credibility of Korea’s AI Trustworthiness technology and verification systems, solidify support systems for domestic industries, and fulfill our role as a specialized institution in the field of AI Trustworthiness.”
Song Sang-hoon, Director-General of ICT Policy at MSIT, emphasized, “The recently passed Korean AI Act in the Legislation and Judiciary Committee provides broad grounds for government support to ensure AI Trustworthiness and safety.” He added, “We will strengthen policy support to secure the trust and safety of the domestic AI industry and actively pursue international standardization in AI Trustworthiness and safety, in collaboration with specialized institutions such as the Korean AI Safety Institute and TTA, to expand the global reach of domestic companies.”
(Reference 1) Link to the NIST website(https://airc.nist.gov/AI_RMF_Knowledge_Base/Crosswalks)
(Reference 2) Link to the Korea AI Safety Research Institute(https://aisi.re.kr/kor/article/ATCLedcbf2856)
|