The use of generative artificial intelligence (AI) in the National Assembly is increasing, prompting calls for systematic security guidelines and management systems. There is a risk of sensitive materials and personal information being leaked if they are input into external AI services, yet there are currently no clear information protection standards, leading to varied judgments among lawmakers' offices. Experts emphasize the need for unified security standards and suggest enhancing the performance of the National Assembly's own AI to reduce reliance on private AI services.
According to a report by Aju Economy on November 11, domestic companies and institutions are implementing AI security solutions to prevent sensitive information leaks. However, the National Assembly's standards and response systems regarding information protection in AI usage remain inadequate. The biggest issue is that decisions related to security, such as whether to input sensitive information, are left to individual lawmakers' offices or staff. One staff member noted, "There are no specific guidelines from the office or the National Assembly when using AI, so we make judgments based on personal standards. While we are cautious with sensitive information, we worry about accountability if issues arise."
Conversely, another staff member argued, "While AI can learn from data, it does not provide specific information to other users," adding, "I’ve heard that Chinese DeepMind is risky, so we only use American AI." Additionally, some officials expressed that using AI poses no issues if the information has been vetted by oversight bodies or is publicly available.
Experts recommend that the National Assembly establish unified AI security guidelines. They suggest defining the types of information that can be input into AI and creating specific security procedures, such as pre-filtering personal or confidential data. Choi Kyung-jin, a professor at Gachon University, stated, "Unless one is an expert, it is difficult to know how to use AI concerning security. We need to set the boundaries for AI usage and establish processes for refining data."
Go Hak-soo, a professor at Seoul National University's Law School and former chair of the Personal Information Protection Commission, emphasized the need for discussions at the National Assembly level to establish principles that individuals can follow and procedures for internal investigations and checks regarding AI usage.
Some also argue for improving the performance and security level of the National Assembly's own AI to reduce dependence on private AI. By creating a secure environment for managing data internally, the risk of sensitive information leaking externally could be minimized.
In April, the National Assembly launched its own generative AI platform, the 'National Assembly AI Legislative Support Platform,' but many aides are either unaware of its existence or do not utilize it. This is attributed to its inferior performance compared to private AI services, making it difficult to use for actual work.
One aide who tried the National Assembly's AI remarked, "The National Assembly's AI analyzes news and social data and presents it as keywords, which is useful for reviewing past events, but it struggles to uncover new information." Another aide added, "The AI's database seems limited to information registered on the National Assembly's website, making it inadequate for work purposes. It would be beneficial if it could evolve into a customized AI that could be used for national audits and other tasks."
In this context, Oh Jeong-ik, a lawyer at the law firm Won AI Data Center, stated, "Ultimately, performance is key. The AI must be capable enough to replace private AIs like ChatGPT for its use in actual work to increase its adoption."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.