Anthropic and South Korea's Ministry of Science and ICT have signed a memorandum of understanding to expand cooperation on AI safety and cybersecurity, formalizing a government-industry partnership as both sides grapple with the risks of increasingly capable AI systems.
Under the agreement, the two parties will collaborate across several areas of advanced AI safety. These include analysis of AI's impact on cyber offense and defense, evaluation of AI model safety and misuse risks specifically in Korean-language contexts, and red-team assessments of autonomous AI agents. The arrangement also covers public-sector AI adoption and information exchange on AI-enabled cyber threats, and involves coordination with the Korea AI Safety Institute.
The Korean-language focus is significant. Most frontier model safety evaluations are conducted primarily in English, which can leave gaps in how systems behave in other languages. By committing to Korean-language evaluations, the partnership aims to surface failure modes and misuse risks that English-centric testing may miss.
The MOU was announced alongside Anthropic's opening of a Seoul office and a wave of enterprise and research partnerships across the Korean AI ecosystem. Together, the moves position the company as a close collaborator with both Korean industry and government at a moment when the two countries' AI relationship is also being shaped by US export controls.
The agreement reflects a broader trend of national governments seeking direct partnerships with leading AI developers to build domestic safety expertise rather than relying solely on external assurances. For Korea, the deal offers access to red-teaming methods and safety evaluation practices; for Anthropic, it deepens relationships with policymakers in a strategically important market.
Officials framed the cooperation as a step toward more rigorous, locally grounded oversight of advanced AI as deployment accelerates across Korean public services and enterprises.
Source: [Digital Watch Observatory](https://dig.watch/updates/anthropic-south-korea-ai-safety-seoul-office)