Government Responds to 'Mythos Security Shock'... "Pushing for Development of AI Security Specialized Models"

Consensus on Seeking Response Measures for High-Performance AI Models
"Establishing Fundamental Defense Systems and Other Countermeasures"

The public and private sectors have joined forces to address the security shock caused by Anthropic's high-performance generative artificial intelligence (AI) model, "Mythos." The government announced that it will push for the development of "AI security specialized models" as part of its efforts to secure AI security sovereignty.

Paik Kyunghoon, Deputy Prime Minister and Minister of Science and ICT, attended the Science and Technology Ministers' Meeting held at the Government Complex Seoul in Jongno-gu, Seoul on the 30th. He spoke on current issues such as the promotion direction of data policy in the AI era of great transformation. 2026.4.30 Photo by Jo Yongjun

On May 8 in Jongno-gu, Seoul, the Ministry of Science and ICT held a roundtable of experts from academia, industry, and research institutes, presided over by Baek Kyun, Deputy Prime Minister and Minister of Science and ICT, to discuss response measures related to global AI companies' cybersecurity projects. The event was attended by companies participating in the "Independent AI Foundation Model Development Project" (Dokpamo), such as SK Telecom, Upstage, and Motif Technologies, as well as major AI companies, academic and industry experts in AI security, and chief information security officers (CISOs) from key companies. Policy chief Choi Woo-hyuk of the Information Security Network Policy Office at the Ministry of Science and ICT told reporters, "As the likelihood of AI-based cyberattacks increases, a systematic response is needed, which is why we convened this meeting."

Participants agreed that it is time to seek both short- and long-term response measures to the security shock triggered by Mythos. During the roundtable, it was reportedly disclosed—based on the consent of a specific company—the details of a cyberattack using Anthropic's generative AI model "Claude Opus 4.7." The demonstration showed how a generative AI model can easily identify cybersecurity vulnerabilities and obtain and access accounts. Policy chief Choi explained, "Although some in the roundtable felt that Mythos was overrated, everyone agreed on the need to seek countermeasures given the emergence of high-performance generative AI models." He added, "Rather than the mere appearance of high-performance AI models, experts pointed out that the greater concern is the potential for large-scale vulnerability discovery. In Korea, the high reliance on legacy systems makes immediate patching difficult."

The need to develop domestic AI security specialized models was also raised as a way to secure AI security sovereignty. Participants emphasized the importance of resolving information asymmetry through engagement in global AI companies' cybersecurity projects and the need for public-private collaboration to respond to AI-driven cyberthreats. Policy chief Choi said, "There was consensus on the need for security-specialized models for generative AI models like Mythos," and added, "We also agreed on the importance of considering how best to utilize outstanding AI systems developed through the Dokpamo project."

Deputy Prime Minister Baek stated, "This issue makes it impossible to further delay a paradigm shift in our society's information security toward AI-based security. We will push for the development of domestic AI security specialized models and, together with experts from various fields, swiftly establish relevant countermeasures, including the spread of the 'Zero Trust' philosophy—which continuously verifies all access across society—and the establishment of fundamental defense systems."

This content was produced with the assistance of AI translation services.