"'Mythos Shock' Requires National Response... Ministry of Science and ICT Prepares Comprehensive Security System"

Actively Seeking Participation in "Project Glasswing"
Guidelines Distributed to Corporate Security Officers

In a swift response to the cyber threats posed by Anthropic’s latest artificial intelligence (AI) model, Mythos, the government is actively seeking to participate in Anthropic’s “Project Glasswing” and is establishing a comprehensive national security framework.

On April 29, the Ministry of Science and ICT announced that, in response to the Mythos shock, it will promptly distribute guidelines outlining security measures that corporate chief security officers and CEOs should implement, applying them in the field as soon as possible. The ministry also plans to provide customized security consulting to small and medium-sized enterprises, which are relatively vulnerable to cyber threats.

These measures were specified following the previous day’s full meeting of the National Assembly’s Science, Technology, Information and Communication Committee, during which lawmakers called for a national response as concerns mounted that Mythos could be abused for hacking. On-site experts emphasized that, with the security threats posed by AI now a reality, the government must accelerate international cooperation, budgeting reforms, and other strategic national responses.

Assemblywoman Lee Juhee of the Democratic Party of Korea stressed, “The question is whether Korea’s security system will function properly in the face of AI threats like Mythos. Dedicated implementation bodies, budgets, professional workforce development, and scaling up AI security must all be integrated into a national strategy.”

Critics also pointed out that as Anthropic expands access rights not only to major U.S. technology companies but also to more than 40 key software organizations, countries and institutions excluded from this list could be left fully exposed to AI-driven hacking threats. Currently, Anthropic is operating Project Glasswing with access limited to U.S. big tech companies such as Apple, Amazon, and Microsoft, as well as select financial organizations, citing safety concerns.

Assemblyman Lee Haemin of the Progressive Innovation Party said, “It is urgent for Korea to be granted access to Mythos,” adding, “Outside the United States, the United Kingdom has secured access through the AI Safety Institute (AISI), but Korea, which aims to become a global top-three AI powerhouse, has not been able to participate in Project Glasswing.”

There was also criticism that the government’s medium- to long-term security technology development plan is set for after 2027, meaning that legal and institutional frameworks are failing to keep pace with technological advances. In response, Ryu Jemyung, Vice Minister of the Ministry of Science and ICT, stated, “We are inquiring about joining Project Glasswing and are already notifying chief information security officers nationwide to raise their security levels.” He added, “It is necessary to reflect the overall strengthening of the security system in next year’s budget, and it is important to take simultaneous action, such as providing consulting to small businesses and distributing guidelines.”

Ryu Jemyung, Second Vice Minister of the Ministry of Science and ICT, is responding to members' questions at the full meeting of the Science, Technology, Information and Broadcasting Committee held at the National Assembly on the 28th. Photo by Yonhap News

Deputy Prime Minister and Minister of Science and ICT Bae Gyeonghun also emphasized that the government is actively pursuing membership in Project Glasswing. After signing a memorandum of understanding (MOU) with Google DeepMind on April 27, Deputy Prime Minister Bae told reporters, “The government is working diligently to participate in Project Glasswing and is actively reaching out to Anthropic regarding participation.”

Deputy Prime Minister Bae also wrote on social media, “AI is fundamentally changing the rules of both cyberattack and defense,” noting, “After assessing the cyber capabilities of the latest models within the range that can be made public, we confirmed that it is possible to identify vulnerabilities and construct attack scenarios at a significant level through prompting alone, without the need for advanced coding.” Highlighting large-scale vulnerability responses, strengthening security inspections of public systems, and participation in Anthropic’s Project Glasswing as short-term priorities, Deputy Prime Minister Bae added, “In the medium and long term, we will establish a system in which AI defends against AI and will accelerate the transition to zero-trust-based security.”

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