AI · 2026-10-15
PASA moves to secure an early lead in international standards for Physical AI
To establish common rules for Physical AI used in manufacturing, robotics, and mobility, the public-private PASA coalition is pushing ahead with domestic standard development and links to international standards. A reference model covering safety, reliability, data management, and cross-sector interoperability has emerged as a key task.
As AI moves beyond processing information on a screen and expands into ‘Physical AI’ that operates in the real world like robots and machinery, the need for common standards that can connect different technologies and industries is also growing. The Physical AI Standardization Forum (PASA), a domestic public-private coalition, has said it will pursue both standards development that reflects industry needs and international standardization.
According to Electronic Times, the Korea Software Industry Association (KOSA) introduced PASA’s direction at a Physical AI standardization strategy seminar held on the 15th at EL Tower in Seocho-gu, Seoul. PASA, operated as part of the Ministry of Science and ICT’s ICT standardization forum support program, is led by KOSA and the Telecommunications Technology Association (TTA). KAIST Professor Jang Young-jae serves as chair, and more than 20 companies and institutions, including Google Korea, HD Hyundai Heavy Industries, and Rebellions, are participating.
PASA plans to identify standardization tasks by surveying technology needs in manufacturing, robotics, and mobility, and to establish a mid- to long-term roadmap. Key development areas include verifying safety and reliability, as well as lifecycle data management. The plan also includes not only establishing domestic standards, but expanding them into global standards through cooperation with international standards organizations.
A common reference model was highlighted as an initial priority. A reference model is a standard that organizes the overall structure and relationships so that multiple technologies and systems can use it as a shared point of reference. Yoo Sang-geun, principal researcher at the Electronics and Telecommunications Research Institute (ETRI), who presented at the seminar, explained that because Physical AI is not limited to a specific industry or a single standards development organization, a common framework covering the whole field is needed. He noted that many existing standards focus on the performance and accuracy of AI models, making it difficult to consistently address the process by which AI judgments turn into physical actions. He also added that standards related to digital twins, operational control, and system lifecycles have each been developed separately, which limits integrated management.
If a reference model is established, it will make it possible to define standards for trust, safety, and accountability, as well as for organizing the meanings of data, models, and actions and enabling systems across different industries to exchange information. That said, standards alone do not guarantee commercialization. Chung-Ang University Professor Park Jun-ho suggested that regulation, standards, and conformity assessment should be linked together alongside technology development. The point is that companies also need to review the purpose of adoption, risk assessment, data integration methods, update and operational responsibilities, and acceptance criteria from the early stages.
For Physical AI to be widely adopted in industrial settings, there must be both a way to verify safe operation and standards that support interoperability between companies. PASA’s standardization work is an effort to build that foundation in Korea and connect it to international discussions. Going forward, a key challenge will be how concretely it can reflect the needs of real-world sites in standards, while also aligning with regulatory and certification systems.
According to Electronic Times, the Korea Software Industry Association (KOSA) introduced PASA’s direction at a Physical AI standardization strategy seminar held on the 15th at EL Tower in Seocho-gu, Seoul. PASA, operated as part of the Ministry of Science and ICT’s ICT standardization forum support program, is led by KOSA and the Telecommunications Technology Association (TTA). KAIST Professor Jang Young-jae serves as chair, and more than 20 companies and institutions, including Google Korea, HD Hyundai Heavy Industries, and Rebellions, are participating.
PASA plans to identify standardization tasks by surveying technology needs in manufacturing, robotics, and mobility, and to establish a mid- to long-term roadmap. Key development areas include verifying safety and reliability, as well as lifecycle data management. The plan also includes not only establishing domestic standards, but expanding them into global standards through cooperation with international standards organizations.
A common reference model was highlighted as an initial priority. A reference model is a standard that organizes the overall structure and relationships so that multiple technologies and systems can use it as a shared point of reference. Yoo Sang-geun, principal researcher at the Electronics and Telecommunications Research Institute (ETRI), who presented at the seminar, explained that because Physical AI is not limited to a specific industry or a single standards development organization, a common framework covering the whole field is needed. He noted that many existing standards focus on the performance and accuracy of AI models, making it difficult to consistently address the process by which AI judgments turn into physical actions. He also added that standards related to digital twins, operational control, and system lifecycles have each been developed separately, which limits integrated management.
If a reference model is established, it will make it possible to define standards for trust, safety, and accountability, as well as for organizing the meanings of data, models, and actions and enabling systems across different industries to exchange information. That said, standards alone do not guarantee commercialization. Chung-Ang University Professor Park Jun-ho suggested that regulation, standards, and conformity assessment should be linked together alongside technology development. The point is that companies also need to review the purpose of adoption, risk assessment, data integration methods, update and operational responsibilities, and acceptance criteria from the early stages.
For Physical AI to be widely adopted in industrial settings, there must be both a way to verify safe operation and standards that support interoperability between companies. PASA’s standardization work is an effort to build that foundation in Korea and connect it to international discussions. Going forward, a key challenge will be how concretely it can reflect the needs of real-world sites in standards, while also aligning with regulatory and certification systems.
Source: “피지컬 AI, 국제표준 선점해야”…PASA, 표준화 추진
Related search terms
#Physical AI standardization#PASA Physical AI forum#AI standards for manufacturing#robotics and mobility standards#reference model for Physical AI#international AI standards#safety and reliability standards for AI