[Kim Kyung-jin Column] Who Will Hold the Line? Data Rules in the Age of Robots



In Liuzhou, Guangxi Zhuang Autonomous Region, China, there is a facility where humans teach robots. Employees wear equipment and repeatedly perform actions such as picking up, moving, and assembling objects. The robots record these actions to learn. According to foreign reports, China has built more than 40 such robot data collection centers by the end of 2025, with 24 currently in operation. Funding for construction has come from both the central and local governments. From 2026 to 2030, China has designated the integration of artificial intelligence with robots, termed 'Physical AI' (or implemented intelligence in Chinese), as one of six key future industries in its five-year plan.

This facility illustrates where competition in the age of robots is determined. Conversational AI like ChatGPT learns from text accumulated on the internet. However, humanoid robots and factory AI face a different situation. There is no material to learn from on the internet. Data such as temperature during welding, the force used to tighten screws, and the condition of equipment at the moment defective products are produced exists only within the factory. The country that secures a large amount of high-quality factory data will lead in the age of robots.

South Korea is a country rich in factory data.

South Korea manufactures semiconductors, automobiles, batteries, ships, and home appliances. Each factory has decades of accumulated process conditions and quality records, and the skills of skilled workers can also be recorded as data. The government is aware of this. The Ministry of Trade, Industry and Energy has formed a Manufacturing AI Transformation (M.AX) alliance with over 1,500 companies and institutions and announced a roadmap for promoting manufacturing data on October 29. Starting in 2027, a national manufacturing data library will be established.

Our companies have factories in the United States, China, Europe, and Southeast Asia. To retrain AI, data from overseas factories must be brought back to headquarters. This is where the problem arises. Each country with factories has different laws regarding data, and these laws often conflict.

China mandates that data be stored within its borders.

China implemented the Cybersecurity Law in 2017. Article 37 of this law stipulates that operators of critical infrastructure such as telecommunications, energy, finance, and transportation must store personal information and 'important data' collected in China within the country. To send data abroad, a government security assessment is required. In 2021, the Data Security Law and the Personal Information Protection Law were added.

Foreign companies are not exempt. In 2018, Apple moved the iCloud data of Chinese users to servers operated by a local company in Guizhou Province. In 2021, Tesla announced it would build a data center in Shanghai and store all data from vehicles sold in China within the country. As it became difficult to bring driving footage collected on Chinese roads to the U.S. for training, Tesla opened a separate AI training center in China this year.

The manufacturing sector is also affected. The guidelines for exporting automotive data issued by the Chinese government this year listed design documents, material lists, production control programs, and autonomous driving training data as important data. The revised Cybersecurity Law, which took effect on January 1 of this year, raised the maximum fine to 10 million yuan. However, in free trade pilot zones such as Shanghai and Guangdong, data not on the export restriction list is exempt from review. This approach maintains control principles while relaxing them only as necessary.

The United States demands free transfer while strengthening security controls.

The United States has been advocating for the free transfer of data to other countries. This is because companies like Google, Amazon, and Microsoft must be able to operate anywhere in the world. The U.S. Trade Representative identified South Korea's public cloud security certification and restrictions on AI-based infrastructure procurement as trade barriers in a trade barrier report released on March 31 of this year. Last year's report criticized our system that prevents national core technology data from being stored on foreign clouds.

Even the U.S. controls data transfer regarding China. A Department of Justice rule effective April 2025 prohibits or restricts transactions that transfer sensitive personal data of Americans to six countries, including China, Russia, Iran, North Korea, Cuba, and Venezuela, and companies controlled by them. While this is not a rule directly governing factory data, it affects data management for companies operating in China. Export controls on advanced semiconductors and manufacturing equipment have also been tightened. The U.S. Department of Commerce has finalized rules prohibiting the export of software and parts for connected vehicles linked to China and Russia starting January 2025. In 2018, the U.S. also enacted the Cloud Act, allowing U.S. law enforcement to request data stored by U.S. companies on overseas servers. The U.S. demands openness from allies while responding to competitors with restrictions.

The European Union requires data to be provided to users.

The European Union has chosen a different direction. The Data Act, effective September 2025, imposes obligations on companies that manufacture connected machines and equipment. They are required to provide data generated from those products to users and third parties. Products released after September 12 of this year must allow users to extract and use data from the design stage. South Korean companies exporting equipment and robots to Europe will also bear this obligation. There is a risk that proprietary technology may leak during the data provision process. The Data Act also requires cloud providers to implement measures to prevent governments outside the EU from accessing industrial data within Europe. The structure allows for data to be provided within Europe while preventing its export outside the region.

Starting February 18, 2027, electric vehicle batteries must have a 'battery passport' to be sold in Europe. This system requires the submission of digital records of the raw material history and carbon emissions of the batteries. To calculate the carbon emissions of finished products, data from component manufacturers will be needed. Small and medium-sized enterprises, including second- and third-tier suppliers, will also be required to submit data.

Data barriers are spreading globally.

The Reserve Bank of India mandated in 2018 that payment data be stored only within India. Vietnam will implement a Data Act in July 2025 to manage the overseas transfer of 'core data' and 'important data.' According to a study by the Information Technology and Innovation Foundation (ITIF), the number of countries restricting data transfer increased from 35 in 2017 to 62 in 2021, and the number of restrictive measures rose from 67 to 144.

Our companies' factories in China are subject to regulations from three regions simultaneously. China reviews data exports, the U.S. controls technology and data going to China, and European clients demand data submission. Companies trying to send records of defect rates from Chinese factories to their South Korean headquarters must first determine whether those records are classified as important data. The criteria for judgment are unclear, making it difficult to predict the outcome of reviews. This is a problem that companies cannot solve alone.

Current international rules do not address factory data.

Inter-country data rules are established through treaties known as digital trade agreements. South Korea signed a digital trade agreement with the European Union in Brussels on June 10. This agreement guarantees the overseas transfer of data, prohibits demands to localize servers, and forbids requests to submit source code, which is the design content of software.

However, the agreements signed so far have been based on online shopping and platform services. Factory data generated by machines and AI models trained on that data are not explicitly mentioned in the provisions. It remains unclear whether there are provisions to prevent foreign governments from requiring the submission of AI learning outcomes (weights) that determine the performance of robots during certification reviews.

As robots become more widespread, the issues will grow larger. Robots working in factories and homes continuously record their surroundings with cameras and sensors. The data collected by foreign robots working in South Korean factories will also become a trade issue regarding which country's server it is sent to in the future.

The provisions requested by South Korea also apply to South Korea.

The reason the solution is not simple is that agreements apply equally to both sides. If South Korea demands the free transfer of data from overseas factories to headquarters, it will also bear the same obligations. This could lead to criticism that our system, which requires government approval for the overseas transfer of national core technologies like semiconductors and batteries, violates the agreement. If we prohibit the submission of AI models, our exporting companies will be protected. However, this would reduce the government's authority to verify foreign AI entering the country.

South Korea is a country that must protect data while also needing data to cross borders for business. It cannot control everything like China, nor can it demand based solely on market size like the U.S. Therefore, sophisticated standards are necessary.

South Korea has four tasks to prepare.

First, data must be categorized. General data, such as equipment temperature, cannot be treated the same as process data that falls under national core technologies. Standards must be established to allow free transfer of general data while strictly protecting core data, which will enable a consistent position in negotiations. The scope of core data should be defined as minimally necessary, and the criteria for judgment should be made public. This way, South Korea can demand the same level of transparency from counterpart countries.

Second, South Korea should propose that factory data and AI models be explicitly included in the agreement provisions. South Korea has a high proportion of manufacturing and served as the secretariat for the International Electrotechnical Commission's Smart Manufacturing Standardization Committee this July. It has the qualifications and capabilities to propose rules in the field of manufacturing data.

Third, cooperation with other manufacturing powerhouses facing similar circumstances is essential. Japan advocates for the principle of freely exchanging data with trusted partners, while Germany operates a data-sharing network in the automotive industry. In October, five domestic companies completed a demonstration exchanging automotive data and carbon emissions data with a European data-sharing network. This cooperation should be developed into intergovernmental agreements.

Fourth, support for small and medium-sized suppliers is crucial. The requirement for data submission from overseas clients ultimately falls on the small and medium-sized enterprises that produce components. These companies should not bear the costs of creating and validating data or the risk of leaking trade secrets alone.

Past trade negotiations dealt with tariffs, rice, and automobiles. Future trade negotiations will focus on data. International rules regarding factory data, once established, will be difficult to change for decades. It is too late to respond after the rules are set. Factory data is an asset for companies and a record of the experiences accumulated by people working in those factories for decades. How that data is used to train robots from which country will directly impact the future of jobs and industries. The government, National Assembly, and businesses must discuss this issue as a national task now.

Author's Major Background

A lawyer who has served as a prosecutor and a member of the National Assembly, as well as a writer in the field of artificial intelligence (AI). He graduated from Korea University with a degree in law and served for about 15 years as a prosecutor and chief prosecutor. As a member of the 20th National Assembly (Gwangju Buk-gap), he engaged in legislative activities in the fields of science, technology, and information and communication. He is researching and writing about the social impact of AI technology and institutional responses based on his legal and policy experience. He also gives lectures on the use of generative AI and research and business innovation to universities and institutions.



* This article has been translated by AI.