The South Korean government is moving forward with its participation in OpenAI's global AI security collaboration framework, known as Trust-Based Cyber Access (TAC). Next week, officials are expected to finalize the scope and operational details of TAC during discussions at OpenAI's headquarters.
According to industry sources, the Ministry of Science and ICT plans to meet with OpenAI next week to clarify the direction of TAC cooperation. A senior official from the ministry stated, "While nothing is confirmed yet, we have reached a broad consensus, and detailed discussions will take place next week with Jason Kwon, OpenAI's Chief Strategy Officer."
During a workshop on OpenAI security held the previous day, both public and private sectors expressed significant agreement on the necessity of TAC collaboration. A ministry official noted, "This workshop was initiated at OpenAI's request, and we view it positively."
Participants in TAC will be granted limited access to OpenAI's security-focused AI models. Currently, global security firms such as CrowdStrike, Palo Alto Networks, Cloudflare, Fortinet, Akamai, and Zscaler are involved in TAC, integrating and utilizing OpenAI models within their security systems.
The government's push for TAC cooperation is seen as a response to the evolving international AI security landscape.
According to the Software Policy Research Institute, the U.S. White House is intensifying efforts to manage AI and cybersecurity as part of its national security strategy. There is a clear trend toward protecting AI models, data centers, and security systems as core assets.
The U.S. is also working on building AI-based threat detection and defense systems, expanding secure AI data centers, internalizing AI security, and establishing frameworks for sharing information on AI vulnerabilities.
Experts have positively assessed the potential of TAC collaboration to enhance South Korea's capabilities in responding to cyber threats. Eom Heung-yeol, a professor at Soonchunhyang University, stated, "By sharing information with OpenAI, we can secure vulnerability data and patch methods, which will help block AI-based attack codes and improve our security response systems."
However, he cautioned against creating a separate organization for responding to new threats, suggesting that enhancing the existing Cyber Threat Information Analysis and Sharing (C-TAS) system, currently operated by the Korea Internet & Security Agency (KISA), would be more practical.
Meanwhile, the government is also in discussions to participate in an AI-based security collaboration program called Project Glasswing with Anthropic. However, given that there are no national-level participation examples and that the only foreign research institute involved is the UK AI Safety Research Institute, actual collaboration may prove challenging.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.