The South Korean government is set to establish a 'Manufacturing Data Library' to utilize core manufacturing data for artificial intelligence (AI) training without disclosing it externally. This initiative aims to address the reluctance of companies to share valuable data, such as information on defects and accidents, due to concerns over trade secret leaks. The government plans to invest 115.8 billion won next year to collect non-public data and develop industry-specific AI foundation models.
Lim Kyung-seop, Director of the Manufacturing AI Transition Cooperation Division at the Ministry of Trade, Industry and Energy, stated on September 3 during the second day of the 18th Good Growth, Good Jobs Global Forum (2026 GGGF) held at The Plaza Hotel in Seoul, "Existing manufacturing data initiatives have struggled because companies were hesitant to share valuable data under public disclosure requirements." He added, "This time, we will fundamentally collect data on a non-public basis, allowing for learning in a clean room, but the data cannot be taken out; only the trained AI models can be exported."
The biggest barrier to AI adoption in manufacturing is data availability. Lim noted, "If we can resolve the data issue, we can consider 80 to 90 percent of the problem solved." According to the ministry, over 90 percent of domestic manufacturing companies are still not utilizing AI. There is also a need for standardization, as companies use different equipment names and measurement units. Lim pointed out, "Without standards, the more data we collect, the more foolish the AI model becomes."
The policy goal has shifted from improving individual processes to developing general-purpose AI. Lim explained, "Previously, we focused on improving a single process within individual factories, but now our goal is to collect more data to create foundation models that can be used across entire processes." The aim is to digitize the tacit knowledge that has relied on skilled workers' experiences in areas like casting and welding, applying it to AI and robotics.
The ministry plans to expand its initiatives centered around the 'M.AX (Manufacturing AI Transition) Alliance,' which was launched last year. Currently, about 1,500 companies and institutions are participating across 11 sectors, including AI factories, robotics, future vehicles, semiconductors, and autonomous ships. The number of AI factories has increased from 102 at the end of last year to approximately 170 processes now. Lim reported that in 38 measurable projects, defect rates have decreased by about 15.5 percent, and productivity has increased by around 30 percent.
Physical AI demonstrations are also set to expand. Humanoids are currently being tested in 10 industrial sites, with plans to increase that number to 30. The development of on-device AI semiconductors is also underway. Lim emphasized, "Autonomous vehicles cannot stop just because communication is lost. We need semiconductors that consume less power, have low latency, and are durable."
The ministry is also working on establishing a legal and institutional framework. It is pushing for the enactment of the so-called 'M.AX Act,' which includes provisions for utilizing manufacturing data and training AI professionals. Additionally, the ministry is working on establishing standards related to manufacturing AI and providing GPU support. Lim noted that the ministry's AI-related budget has increased from approximately 1.2 trillion won this year to 1.7 trillion won next year, stating, "We will actively promote budget and funding support centered on key projects."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.