The AI Team at Bae, Kim & Lee LLC is publishing a series of newsletters introducing and analyzing the subordinate guidelines released by the Ministry of Science and ICT (“MSIT”) under the AI Basic Act. The third topic in this series is “Determining High-Impact Artificial Intelligence.” In this issue, we review the key contents and implications of the Guidelines for Determining High-Impact Artificial Intelligence (the “Guidelines”).
I. DEFINITION AND CRITERIA FOR DETERMINING HIGH-IMPACT ARTIFICIAL INTELLIGENCE
High-impact artificial intelligence refers to an artificial intelligence system that (i) is likely to have a significant impact on, or pose risks to, human life, physical safety, or fundamental rights, and (ii) is utilized in specific areas (see I.(1) below) (Subparagraph 4 of Article 2 of the AI Basic Act on the Promotion of AI Development and Establishment of Trust (the “AI Basic Act”)).
If an AI system falls under the category of high-impact artificial intelligence, any person intending to provide products or services using such AI must notify users in advance that the services are operated based on high-impact AI and must implement measures to ensure reliability and safety. Accordingly, accurately determining whether an AI system qualifies as high impact is of considerable importance.
1. STEP 1 CRITERIA
The AI system must be used in one of the following areas:
2. STEP 2 CRITERIA
An artificial intelligence system used in one of the areas identified in Step 1 will be classified as high-impact AI if it meets the Step 2 criterion of being “likely to have a significant impact on, or pose risks to, the protection of human life, physical safety, or fundamental rights.”
In assessing the significance of impact and risks, the intended purpose, functions, and usage context of the AI system must be examined from multiple perspectives. In particular, the significance of risks should be determined through a comprehensive assessment of (i) the essential nature of the rights and interests at stake, and (ii) the extent to which the risk is heightened.
For this purpose, it is desirable to use measurement methods and indicators in risk assessments; however, rather than relying entirely on such indicators, balanced consideration should be given to the AI system’s purpose, functions, and usage context, including positive effects and social acceptability.
II. RIGHTS AND INTERESTS BY AREA AND EXAMPLES OF SIGNIFICANT IMPACTS OR RISKS CAUSED BY AI
The key criteria for determining high-impact AI by sector, along with relevant and non-relevant examples provided in the Guidelines, are summarized below.
III. IMPLICATIONS AND NEXT STEPS
The Guidelines suggest that companies utilizing artificial intelligence should take the following actions:
First, conduct a company-wide review and self-assessment of AI usage. Companies should establish procedures to internally assess whether the AI systems used in their products or services fall under the category of high-impact AI according to the detailed criteria set out in the Guidelines. Where it is difficult to make such a determination internally, companies may request the MSIT to confirm whether their AI qualifies as high-impact AI (Article 33 of the AI Basic Act).
Second, establish a compliance framework. If classified as a business operator utilizing high-impact AI, a company must implement internal policies and processes to fulfill obligations such as:
-
formulating and implementing risk management and user protection measures,
-
ensuring explainability of AI outputs and training data, and
-
incorporating human intervention where necessary.
The company must also develop the capacity to systematically document and manage these matters (Article 34 of the AI Basic Act). For more detailed requirements, please refer to the Guidelines on the Responsibilities of High-Impact AI Operators, which were released together with the Guidelines.
* * *
BKL’s AI Team continues to monitor domestic and international regulatory developments concerning artificial intelligence. Leveraging its expertise, the team provides comprehensive advisory and practical support to help companies deploy AI technologies safely while minimizing legal risks. Please feel free to contact us with any inquiries.
[Korean version]
