KAIST has become the only institution in South Korea selected for Microsoft’s global artificial intelligence (AI) safety research program, aiming to develop next-generation AI security and reliability technologies.
On July 28, KAIST announced that Professor Son Su-el from the Department of Computer Science has been chosen as a research support recipient for the EXTRA (External Red Team Alliance) program, operated by Microsoft’s AI Red Team.
EXTRA is a newly launched global research support initiative by Microsoft’s AI Red Team. It aims to validate the safety and security vulnerabilities of AI systems from various perspectives and to expand an independent AI safety research ecosystem. This program is significant as it broadens AI safety assessments beyond internal corporate evaluations to include universities and research institutions worldwide.
Microsoft determined that as AI systems become more advanced, it is essential to comprehensively validate expertise across various fields, including cybersecurity, multilingual environments, regional and cultural contexts, AI alignment, and potential misuse. They concluded that a single organization could not adequately assess these risks, leading to the establishment of EXTRA.
As part of this selection, KAIST will receive a research grant of $25,000 (approximately 37 million won) from Microsoft for research on AI safety, security, AI alignment, and responsible AI development. This funding will be provided in the form of a 'gift award,' with no specific timeline for the research project, and will be utilized by Professor Son’s research team for their AI security and safety studies.
EXTRA includes participation from over ten universities across six continents, in addition to KAIST. Microsoft plans to use this program to expand an independent AI safety research ecosystem and strengthen collaboration between academia and industry.
Professor Son stated, "As AI systems rapidly spread across society, ensuring not only performance but also safety and reliability has become paramount. Through our participation in Microsoft’s AI Red Team EXTRA, we aim to systematically analyze various security threats that may arise in real AI service environments and contribute to the development of AI technologies that everyone can use with confidence."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.