Chris Choury, head of Anthropic International, said this at a press briefing on Jun. 17 at the Conrad Hotel in Yeouido to mark the opening of Anthropic's Seoul office. He dismissed market concerns following the U.S. government's move to impose export controls that block foreign access to Anthropic's cutting-edge AI models "Mythos 5" and "Fable 5."
He said, "The 'jailbreak' case cited as the backdrop for the export controls has also been found in the vast majority of AI models released over the past six months," adding, "We expect to be able to use the Mythos 5 and Fable 5 models again within a few days." Anthropic dispatched a delegation of top technical leaders to Washington over the weekend to enter talks with the Donald Trump administration over AI export controls, signaling that a smooth agreement is likely soon.
"Mythos 5," whose use was halted by White House regulations, is the enterprise and institution service version of the high-performance AI model "Mythos Preview," which offers expert-level security vulnerability detection. "Fable 5" is a version refined for general users, equipped with safeguards to prevent hacking and misuse. Earlier, Anthropic launched "Project Glasswing," allowing Mythos access only to select institutions due to the risk that "Mythos" could be misused for hacking and the like.
The Washington Post (WP) recently reported that the U.S. government moved to impose an export ban on Anthropic's latest AI models because the White House found a "Korean telecom company" suspected of ties to China among 111 corporations granted priority access to "Mythos 5," sparking controversy. Among Korean corporations, Samsung Electronics, SK hynix and SK Telecom newly joined Project Glasswing early this month.
However, on the controversy over "Project Glasswing," Choury offered only a basic response, saying, "We have nothing to share at this time."
Anthropic said it officially opened its Seoul office that day and will ramp up its business in Korea. Choury said, "Korea is currently one of the fastest-growing markets in the world," adding, "Korea ranks 12th globally in per-capita 'Claude' usage among 116 countries, and considering Korea's strong tech talent and developer ecosystem, it is highly likely to move into the single digits soon." In fact, according to an economic index report Anthropic released in March, Claude usage in Korea exceeded expectations based on population size by 3.5 times.
Choury did not disclose specific growth rates for the Korean market but said it is tracking Anthropic's global growth trend. He said, "Anthropic's annual recurring revenue (ARR) increased from $9.0 billion (about 13 trillion won) at the end of last year to $47.0 billion (about 71 trillion won) as of last month's investment round, and we are seeing this growth in the Korean market as well."
Anthropic plans to expand cooperation across Korea's AI ecosystem, spanning major corporations, startups and AI research institutions. As Anthropic accelerates its push into the Korean market for corporations, competition with OpenAI, the developer of ChatGPT that entered earlier last year, is expected to intensify.
Asked about differentiation from rival OpenAI, Choury highlighted the technical edge of the company's AI models and its strong position in the market for corporations, saying, "From its founding, Anthropic has prioritized AI safety above all and focused on the enterprise market."
He said, "'Opus 4.8,' which Anthropic launched a few weeks ago, is the highest-performing model at this point based on major benchmarks, and the 'Mythos Preview' unveiled in early April is our most powerful model, which early customers have rated as having security capabilities on another level," expressing confidence in its AI model competitiveness. He added, "Anthropic's market share in AI for corporations was 40% as of January this year, and it is likely higher if we reflect the latest figures."
In Korea, major corporations such as Naver, LG CNS, Nexon and Hanwha Solutions are using Anthropic's flagship AI coding agent "Claude Code" and the corporate AI service "Claude Enterprise."
Naver recently rolled out Claude Code across its entire engineering organization. Thousands of developers are using Claude Code, which Anthropic said is the largest deployment in Asia. Nexon's engineering organization is also using Claude Code for game code authoring, review and deployment.
Among large corporations, LG CNS signed a contract earlier this month with Anthropic to adopt "Claude Enterprise." LG CNS opened Claude to all employees for a variety of tasks, including software development, and plans to expand Claude adoption across LG Group going forward. Hanwha Solutions is providing Claude to employees through AWS Bedrock, and Samsung SDS is rolling out Claude for Samsung Electronics employees.
In addition, it has partnered with AI startup Wrtn Technologies, legal tech startup Law&Company and Channel Corporation.
Choi Gi-young, Anthropic's Korea head, said, "To support domestic corporate customers, we will form dedicated teams with expertise in each industry and continuously enhance Claude's Korean performance and user experience." He added, "We will also work closely with cloud service corporations (CSPs) such as Amazon Web Services (AWS), Google Cloud and Microsoft, and with domestic system integrator (SI) corporations including Samsung SDS and LG CNS, to build a stable partner ecosystem."
Anthropic said that to reliably supply AI models to corporate customers, it is working with hyperscalers such as AWS, Google Cloud and Microsoft, which operate large-scale data centers, while also reviewing various options to build its own infrastructure. Choury said, "While we cannot disclose specifics, we are in discussions with major semiconductor corporations in Korea."
Anthropic is an AI corporation founded in 2021 by seven former OpenAI researchers, including Chief Executive Officer (CEO) Dario Amodei. It has grown rapidly by focusing on the corporate market around its flagship AI model, "Claude."
Ahead of an initial public offering (IPO) this fall, the company raised $65.0 billion (about 97 trillion won) in a Series H funding round last month. With this fundraising, its valuation was assessed at $965.0 billion (about 1,445 trillion won), surpassing for the first time rival OpenAI's valuation of $852.0 billion.