Home > Business > Tech

Korean tech giants in talks with gov't to join Anthropic's Project Glasswing in face of 'Mythos shock'

Published: 29 May. 2026, 05:00

Silhouettes of laptop users are seen next to a screen projection of binary code in this illustration created on March 28, 2018. [REUTERS]

Korea’s major tech firms are in discussions with the government about joining Project Glasswing, a cybersecurity initiative launched by U.S. AI giant Anthropic in response to the potentially disruptive capabilities of its own AI model, Mythos.

The consultations come as the country became the latest to join OpenAI's Government and Institutional Trust-Based Access Program alongside Japan, which specializes in AI-powered cyber defense.

“We are reviewing various measures to build a joint response framework for frontier AI models like Mythos, including discussions with government agencies and participation in domestic and overseas AI security cooperation projects such as Project Glasswing,” said a source at one of Korea's major tech companies, who spoke on the condition of anonymity as the discussions are ongoing.

Ryu Je-myung, second vice minister of the Ministry of Science and ICT, far right, meets with OpenAI officials to discuss cooperation in responding to AI security threats at the Three IFC building in Yeouido, western Seoul, on May 26. [NEWS1]

Mythos can detect flaws in what had been believed to be unbreakable code, meaning that in theory, a single prompt by a bad actor could have devastating repercussions. Anthropic claims that the impact could be so immense that it withheld Mythos from public access, limiting access only to select participants under the Glasswing network.

Out of 1,000 open-source projects, which according to Anthropic “collectively underpin much of the internet,” the model found what it estimated were 6,202 high-or critical-severity vulnerabilities in these projects.

Anthropic is accelerating its entry into Korea with the appointment of Choi Ki-young as Representative Director of Korea, a sign of increased focus and the growing popularity of their pre-existing Claude model in the Korean market.

The Anthropic logo is seen in this illustration created on March 1. [REUTERS]

Korean tech wants in
Industry insiders project that interested firms may include Samsung Electronics, LG Electronics and SK hynix as well as financial firms, as their massive trove of intellectual property and privacy-related data could become subject to the growing security threats presented by advanced AI systems.

The logo of Samsung Electronics is seen on a gate at the company's headquarters in Suwon, Gyeonggi, on May 22. [AP]

Project Glasswing currently has approximately 50 partners, according to Anthropic, including firms such as Google, Nvidia, Microsoft and Apple. No Korean companies are currently included, according to multiple industry sources.
The Japanese government is already scheduled to obtain Mythos in late May, followed closely by its megabanks Mitsubishi UFJ ‌Financial ⁠Group, Mizuho Financial Group and Sumitomo Mitsui Financial Group, while Korean tech and finance firms are still in limbo.

Still, the Korean government has so far adopted a measured approach despite calls from companies seeking quicker government backing for participation in the Anthropic project.

“Past large-scale hacking incidents have shown that cybersecurity risks can no longer be confined to individual companies and may quickly escalate into national security concerns, underscoring the urgency for the government to establish AI-driven cyber threat responses, including potential participation in Project Glasswing,” the source said.

An executive at a major Korean chipmaker raised the need for the government to play a leading role.

“There are clear limits to how individual companies can respond on their own. Our responses are confined to monitoring of the news across the globe. Close coordination with the government is essential,” the executive said.

The logo of SK Hynix is seen on one of its products during the 26th Semiconductor Exhibition in Seoul in 2024. [Reuters]

A spokesperson from the Science Ministry told the Korea JoongAng Daily there were no regular meetings between the government and Korean tech companies to consult on discussions with Anthropic and OpenAI.

Korea vulnerable to AI weaponization
Kim Seung-joo, a professor of cybersecurity at Korea University, said Korea is “structurally more vulnerable” to what he called a “Mythos shock” while speaking at a seminar focused on AI defense on May 13 at Korea University.

“Technology is not the problem,” said Prof. Kim. “We can acquire the technology. The problem is that the ecosystem and culture needed to actually utilize such technology is where Korea is more vulnerable.”

Prof. Kim was referring to the lengthy chain of regulatory frameworks that slowed the government's response to cybersecurity issues. With a single malicious prompt, Mythos could potentially bring down large-scale security apparatuses in both public and private sectors within just 10 hours.

Korea's financial sector is especially susceptible because it adopts a network separation rule as part of its current cybersecurity approach. Internal corporate networks are segregated from public internet networks, meaning that inside companies, people cannot use third-party AI models, which hampers their ability to respond to digital threats.

A spokesperson at Hana Bank said AI tools such as Mythos cannot be put into action even if Anthropic granted access, because government regulation comes first, echoing Prof. Kim's remarks.

“In Korea, banks first need regulatory approval before using these kinds of systems. The domestic financial sector operates under a licensing and approval framework.

That means we need to comprehensively review all related restrictions and regulatory considerations. So we are not really at a stage where we can definitively say yes or no. It’s not at the stage where the bank can officially discuss plans one way or the other.”

A currency trader watches monitors near a screen showing the Korea Composite Stock Price Index (Kospi) at the foreign exchange trading room of Hana Bank's headquarters in Jung District, central Seoul, on July 9, 2025. [AP]

The spokesperson said “extensive discussions” generally take place with regulators and government authorities first. He said the financial industry was “unique” in that regard.

Prof. Kim said Korea's financial sector was the most inflexible to rapid change and response, after government and public sectors. “The private sector, including telecoms, is equally inflexible,” said Kim.

Korea and Japan are home to some of the largest tech and banking sectors in the world outside of the United States, making digital defense paramount in an AI boom.

Governmental response lags
Science Minister Bae Kyung-hoon, together with Foreign Affairs Minister Cho Hyun, National Intelligence Service Director Lee Jong-seok and other agency heads met on May 11 to “explore avenues for cooperation with Anthropic in the areas of artificial intelligence and cybersecurity — with the aim of securing AI safety and trust and promoting the domestic AI industry,” according to a statement released by the Science Ministry.

Government officials “engaged in in-depth discussions on various cooperation topics” with Anthropic representatives, including Michael Sellitto, head of global affairs.

A Science Ministry Spokesperson said the ministry could not disclose further discussions around access to Project Glasswing, only that there would be an announcement “if Korea decides to come on board.”

OpenAI CEO Sam Altman attends a session with SoftBank group Chairman and CEO Masayoshi Son in Tokyo on February 3, 2025. [AFP]

The Science Ministry has gone on to add that Mythos-specific measures would be announced sometime in the near future.

Japan and Korea are unique in the fact that, outside of the United States, they house some of the biggest tech firms that are also globally dominant in their industries. Because being part of Project Glasswing acts almost like being on a “world's richest” list, these firms feel they’ve earned a spot.

The Japanese government and banking sector have seemingly been granted access to Mythos, even as Korea held largely similar talks with Anthropic's executives days before that did not reach the same result. Korea's regulatory rigidity seems to be the reason for the delay.

The Japanese Financial Services Agency (FSA) held a working group meeting on April 24.

The purpose of the working group was to “deepen discussions to ensure that the financial industry, IT service providers and relevant public bodies — including the government and the Bank of Japan — share a common understanding of threats arising from advances in AI technology and jointly consider appropriate responses,” according to a statement from the FSA.

While Korea's neighbor makes progress, its own cybersecurity environment remains structurally vulnerable, lacking not only a comprehensive legal framework but also sophisticated systems.

“Korea lags behind major economies in areas such as the absence of a comprehensive cybersecurity law, with the lack of advanced systems that automatically match cyberthreats with affected IT assets and mandatory SBOMs [software bills of materials],” said Lee Sang-kyun, head of the AI Security Research Institute at Korea University, during a seminar hosted by the National Assembly Research Service.

BY FERGUS GOODALL SMITH, PARK EUN-JEE [[email protected]]

Ai Mythos Government Claude OpenAI Cybersecurity

More in Tech

Even Sonny's impressed as Hyundai's Atlas robot shows off trick 'Ghost Rabona' in football-themed campaign

SK Group chairman says most valuable people in AI era will be those who connect humans and technology

'I'm digging my own grave': The AI push in workplaces is causing blues and burnout

Korean tech giants in talks with gov't to join Anthropic's Project Glasswing in face of 'Mythos shock'

Naver to pump 1 trillion won into content to power AI evolution

korea joongAng daily

Home > Business > Tech

Korean tech giants in talks with gov't to join Anthropic's Project Glasswing in face of 'Mythos shock'

Related Article

Cybersecurity in 'turbulent transition period' as AI changes game, KAIST professor says

The butterfly effect of the Anthropic contract termination

Rise of AI raises fears of North Korean hacking capabilities

More in Tech

Related Stories

You might also like