Seoul's bid for Anthropic's Mythos AI access hits wall of silence in Washington

South Korea's government contacted the White House to secure access to Anthropic's Mythos AI model, but the effort has produced no tangible results, it has been confirmed. The Ministry of Foreign Affairs has been tracking related reports at the ministerial level, yet progress on obtaining Mythos access remains stalled.

Mythos is an AI model that has advanced to the point where it can autonomously identify security vulnerabilities and carry out hacking on its own.

With the US government offering no clear position on the access restrictions, concerns are growing over AI-driven cybersecurity threats.

Voices in the National Assembly have also warned that "given the possibility that even existing access to frontier-class models could be curtailed depending on further US government action, the government must urgently prepare countermeasures."

According to a document titled "Current Status of Mythos Access" obtained through the office of People Power Party lawmaker Kim Chang-gyum, Ambassador to the United States Kang Kyung-wha held a video conference with the White House's National Cyber Director on June 18 to formally request access to Mythos.

The diplomatic push did not stop there. The Ministry of Foreign Affairs separately tracked multiple Mythos-related reports, and the Ministry of Science and ICT has been holding ongoing consultations with Anthropic.

The US government had earlier imposed export controls covering all Mythos models, including versions 5.0 and 5.1. In June, Washington issued an export control order against Mythos 5 and the general-purpose Fable 5, citing concerns over "technical jailbreaking." Although an easing of those controls was announced later that month — with Fable 5 made publicly available — Mythos 5 remains under US government control. Mythos 5.1, released Tuesday, is subject to the same restrictions.

The US government has said it is "working to resolve the issue expeditiously," but has yet to clarify its position on restricting South Korean companies' access to Mythos. As a result, the Korea Internet & Security Agency, SK Telecom, SK hynix, Samsung Electronics and others have all been unable to use the model.

South Korea's government has responded by pursuing alternative channels. Facing difficulty securing Mythos access, it obtained access to OpenAI's GTAC in late May and Google's Gemini Flash Cyber in late July. More recently, it applied for access to Microsoft's Mdash.

The Ministry of Science and ICT said it "currently does not have secured access rights to Anthropic's Mythos and similar models" and that "the government will obtain access to models with performance comparable to Mythos and use them to assess vulnerabilities in infrastructure, businesses and other entities."

Experts, however, argue that securing Mythos access should be viewed not merely as a matter of obtaining a usage license, but as a question of enhancing national security capabilities.

Without guaranteed access to Mythos, designing a defensive framework capable of countering AI-based cyber threats becomes fundamentally difficult, experts say. An industry official who requested anonymity stressed that "South Korea also needs to directly test and analyze the Mythos model," adding that doing so "would allow us to set a benchmark for how far we need to develop our own frontier models and security-specialized foundation models, and would give us the ability to proactively respond to new AI-based attacks."

In the National Assembly, lawmakers expressed concern that the government has been unable to even determine the US government's true intentions behind the Mythos access restrictions.

People Power Party lawmaker Kim Chang-gyum said the government "cannot rule out the possibility that access to the frontier AI models it currently holds could also be restricted depending on US government policy," adding that "the government has not even been able to properly determine why the United States restricted access to Mythos in the first place."

Kim also said he was worried that "signs of strain in the South Korea-US alliance may be affecting areas including defense capabilities, cybersecurity and the AI industry."