Anthropic Opens Seoul Office While US Export Controls Block Its Top Models in Korea

June 28, 2026 · 31 min read 한국어 원문

On June 17, 2026, Anthropic officially opened its Seoul office — its third base of operations in the Asia-Pacific region, following Tokyo and Bengaluru. The opening day brought a wave of announcements: major Korean enterprises including Naver, Samsung SDS, LG CNS, Nexon, and Hanwha Solutions all signed partnerships with the AI company. Yet just days earlier, the US government had imposed export controls that effectively blocked Korean users from accessing Anthropic’s most advanced models. It was, in effect, a welcome party where the invitations were revoked at the door.

Why Seoul Became Anthropic’s Third APAC Hub

Anthropic describes Korea as “one of the fastest AI-adopting markets in the world” — and the numbers back that up. Claude.ai usage in Korea, measured on a per-capita basis, runs at 3.5 times the global average, placing the country among Anthropic’s top 12 markets worldwide. Demand for Claude-powered automation has grown rapidly, especially among developer communities and enterprise IT departments, which led Anthropic to choose Seoul as its third APAC hub after Tokyo and Bengaluru.

The Seoul office is led by KiYoung Choi, a tech business executive with 30 years of industry experience. Anthropic has signaled its intention to collaborate across three pillars of Korea’s AI ecosystem: enterprise, research, and nonprofit. On the government side, Anthropic signed a memorandum of understanding with South Korea’s Ministry of Science and ICT (MSIT) to cooperate on public-sector AI safety evaluations and cybersecurity initiatives. The company also announced a research support program that will provide free access to Claude models for 60 researchers at institutions belonging to the National AI Research Lab (NAIRL) consortium, including KAIST, Korea University, Yonsei University, and POSTECH.

white and brown concrete buildings during daytime Photo by katherine medelo on Unsplash

Why Naver, Samsung, and LG Are Betting on Claude

The list of partnerships announced at the Seoul opening is striking in both scale and depth.

Naver — South Korea’s dominant search and technology conglomerate — has rolled out Claude Code across its entire engineering organization, making it the largest enterprise Claude Code deployment in Asia, according to Anthropic. Thousands of Naver developers are now using AI for code generation, review, and workflow automation.

Samsung SDS, the IT services arm of the Samsung Group, is deploying Claude Cowork and Claude Code to Samsung Electronics employees, focusing on knowledge-work automation and large-scale software development. LG CNS, the tech subsidiary of the LG conglomerate, is simultaneously expanding Claude usage across LG’s business units and building Claude-powered solutions for its corporate clients. Game developer Nexon is applying Claude Code to live-service game development pipelines, and energy company Hanwha Solutions is using Claude via AWS Bedrock — a setup that allows it to keep data processing within Korea’s borders.

The common thread across all these adoptions is performance on coding and agentic tasks. Claude Code has posted strong scores on SWE-bench, the widely used software engineering benchmark, and has quickly established itself as a serious productivity tool for developers. Taken together, these deals illustrate a broader shift: Korean enterprises are now treating AI not as a chatbot layer, but as core operational infrastructure.

Modern skyscrapers against a cloudy sky Photo by DaVith on Unsplash

The Paradox: Export Controls Arrived Five Days Before the Ribbon-Cutting

On June 12 — five days before the Seoul office opened — the US government issued export controls blocking foreign access to Anthropic’s two most powerful models: Fable 5 and Mythos 5. The stated rationale was national security.

The trigger, as reported by US media, was specific. SK Telecom, South Korea’s largest mobile carrier, had gained access to Mythos 5 through Project Glasswing — an invitation-only cybersecurity consortium run by Anthropic. US authorities reportedly flagged concerns about SK Telecom’s alleged connections to China and brought these to the White House’s attention. SK Telecom has denied any wrongdoing, and no supporting evidence has been made public. Separately, Amazon — Anthropic’s largest investor — reportedly identified an instance of safety guardrail circumvention involving Fable 5 and notified the White House, a development that further accelerated the export control decision. The result was a blanket revocation of top-model access for Korean institutions across the board.

The day after the Seoul office opened, Anthropic’s head of international business Chris Ciauri expressed confidence that access would be restored “within a few days.” That optimism proved premature. On June 21, NSA testimony before Congress extended the controls further. As of the time of writing, Korean users remain locked out of Anthropic’s frontier models.

This situation is not isolated to Anthropic. On June 2, the Trump administration issued an executive order titled “Promoting Advanced AI Innovation and Security,” establishing a framework for federal government pre-access review and voluntary evaluation of frontier AI model development and deployment. The policy direction is clear: the US government increasingly treats advanced AI as a strategic national security asset subject to export discipline.

gray concrete building during daytime Photo by Felix on Unsplash

AI Adoption Speed vs. Technological Sovereignty

The implications of this episode extend well beyond a temporary model access blackout. As Korean companies like Naver, Samsung, and LG move quickly to embed foreign AI into their core development infrastructure, the ability to access that technology has proven vulnerable to a single policy decision made in Washington. That structural dependency is now impossible to ignore.

To be clear, the current access block will likely be resolved. Negotiations between Anthropic and US officials are ongoing, and given Anthropic’s strategic investment in Korea, a prolonged total blackout seems unlikely. But the episode has established a precedent: this can happen again, at any time, for reasons that may be only partially disclosed.

Anthropic’s arrival in Seoul is genuinely good news for Korea’s AI ecosystem. The partnerships are real, the research collaborations are meaningful, and the commercial momentum is substantial. But the same week also laid bare how deeply Korea’s AI growth is entangled with the decisions of US technology companies and the US government.

The gap between Korea’s ability to rapidly adopt cutting-edge AI and its ability to exercise autonomous control over that technology is the real question Anthropic’s Seoul office leaves behind.

Sources

Anthropic Opens Seoul Office While US Export Controls Block Its Top Models in Korea

Why Seoul Became Anthropic’s Third APAC Hub

Why Naver, Samsung, and LG Are Betting on Claude

The Paradox: Export Controls Arrived Five Days Before the Ribbon-Cutting

AI Adoption Speed vs. Technological Sovereignty

댓글