▲ Anthropic
The U.S. administration has imposed export controls on "Mythos," the flagship AI model from Anthropic, effectively blocking access for South Korean government agencies and companies that had recently joined the global AI security alliance, "Project Glasswing."
Anthropic announced on June 12 (local time) that the U.S. government has issued export control guidelines, based on directives from national security authorities, that completely suspend access to "Mythos 5" and "Fable 5" for foreign nationals.
The measure applies not only to overseas users but also to foreign nationals residing in the United States and even foreign employees working at Anthropic.
Project Glasswing is a global cybersecurity cooperative launched by Anthropic alongside the unveiling of its Mythos model.
To prevent the model from being misused by malicious hackers, the project has operated by providing early access to verified companies and institutions to help identify and defend against software vulnerabilities.
On June 2 (local time), Anthropic significantly expanded the scope of Project Glasswing, adding approximately 150 new institutions across 15 countries.
In South Korea, Samsung Electronics, SK Hynix, SK Telecom, and the Korea Internet & Security Agency (KISA) were among those reported to have joined.
While these South Korean entities gained access to Mythos upon joining the alliance, marking a first step in global AI security cooperation, they have been hit by the U.S. administration's export control measures just about ten days later.
It is understood that while they were granted access, they had not yet reached the stage of utilizing the model in a practical capacity.
An official from the Ministry of Science and ICT stated, "As we have been in continuous communication with Anthropic, we are currently calmly assessing the facts," adding, "We are discussing response measures through a consultative body centered around the National Security Office."
The government previously announced a "Private Information Protection Promotion Plan to Respond to AI-based Cyber Threats" at the Science and Technology Ministerial Meeting on May 29.
At that meeting, the government stated that it would establish an emergency system centered on the National Security Office to rapidly share and disseminate information on AI vulnerabilities and enable joint responses in the event of security breaches, while also activating a general situation room within the Ministry of Science and ICT along with situation rooms for each relevant ministry in the private sector.
South Korean companies are also reportedly communicating with Anthropic to grasp the situation.
The incident has reignited discussions in South Korea regarding the importance of securing independent AI capabilities, often referred to as "Sovereign AI."
Ha Jung-woo, former AI Future Planning Secretary to the President, assessed the situation immediately after the U.S. administration's export control guidelines were issued, writing on his personal social media, "This is what happens when you become dependent on AI technology in an era where AI capabilities determine national competitiveness."
He emphasized, "Such incidents can continue to happen at any time," adding, "While we must cooperate globally, we must also possess our own capabilities for contingencies."
Regarding the U.S. administration's move, Anthropic has stated that it believes the issue stems from a misunderstanding and intends to restore services as quickly as possible.
Industry observers suggest that access restrictions could remain in place for several weeks until the U.S. government's security framework is reinforced.
(Photo: Yonhap News)
※ Please note: This article was translated by AI and may contain errors.