The safe coexistence of humans and robots has emerged as a new focus in South Korea, which is aiming for global leadership in the humanoid robot industry. Calls are growing for a safety certification system that would allow companies and consumers to use humanoids with peace of mind. Analysts say that building a standards-based physical AI society that ensures safety in advance could also contribute to overseas exports.
According to the information technology (IT) industry on Monday, the Ministry of Trade, Industry and Energy announced an implementation plan in March this year for a project to build a humanoid robot safety certification center. The aim is to establish a hub to support safety and security certification of domestic humanoid robots. The center is expected to be located in the Daegu-Gyeongbuk region, in line with the direction of the three mega-projects the government recently announced.
The center is expected to serve as a testbed for evaluating and demonstrating not only the safety of humanoid robots but also the artificial intelligence (AI) models applied to the robots and their security levels. To this end, dynamic stability evaluation equipment will be installed to verify robots' walking, stopping posture, and walking support force distribution. A reliability evaluation system for AI models and a verification framework for communication security vulnerabilities will also be built.
The government has taken these steps because it is urgent to establish safety and security verification and certification systems tailored to different usage environments for humanoid robots, which are expected to raise various safety concerns. The goal is to prevent risks in advance, such as production lines being halted by AI robot accidents or people being injured by malfunctions. Kim Su-hyung, a department head at the Electronics and Telecommunications Research Institute (ETRI), said at a recent forum, "One security AI technology detected 38 vulnerabilities even when it knew only the product names of three types of robots," adding, "Safety and security must be ensured across all layers of physical AI, including recognition, planning, judgment, execution, and control."
Safety certification is also considered important from an export perspective. Because humanoid robots are an assembly of various technologies such as AI, components, and communications, they must meet multiple international certifications. Song Chang-jong, head of the Device AX Innovation Team at the Ministry of Science and ICT, explained, "There is still no comprehensive standard worldwide to objectively verify the safety and reliability of physical AI," adding, "We are trying to secure standards first through various means, including global cooperation."
Analysts also say that a reliable certification system is essential to expand adoption in the domestic industry. In fact, collaborative robots were once subject to regulations requiring fences to be installed in the mid-2010s due to the absence of a safety verification body, which led to sluggish adoption for several years, especially among small and medium-sized enterprises. Taking this as a cautionary lesson, they say that a systematic humanoid safety certification system could prevent the side effect of legal regulations delaying market growth. In particular, as the government has said it will commercialize humanoid robots specialized for 10 key industries by 2028 and deploy 1,000 AI robots to worksites annually, there is an urgent need to create conditions for humanoid robots to work safely from the early stages of the industry.
Experts advise that the humanoid robot certification system should be organized in a way that covers the entire lifecycle, from before launch to after deployment. Cha Nam-jun, a senior researcher at ETRI, said in a report titled "Major Regulatory Trends and Recommendations in the Physical AI Field: Focusing on the US, EU, and China," "Physical AI regulations in countries such as the US, the European Union (EU), and China are provided in the form of management systems spanning the entire lifecycle of the system," and recommended, "The government should recognize technological innovation and risk prevention not as separate tasks but as a single policy goal, and concentrate its policy capabilities on proactively building a trustworthy technological environment." Kim Jung, a professor at the Korea Advanced Institute of Science and Technology (KAIST), assessed, "Above all, it is important to accumulate successful cases of mass-producing humanoid robots."
