The Volunteer Agency for Korean Image (VANK), a Seoul-based civic organization, has formally called for the development of sovereign artificial intelligence (AI) systems in South Korea following reports that major generative AI platforms are restricting or misrepresenting historical data regarding the Liancourt Rocks, known as Dokdo in Korea. The organization argues that reliance on foreign-developed AI models poses a significant risk to the accurate documentation of Korean sovereignty and history.
The controversy centers on the output generated by multinational AI services, including concerns raised by users regarding Google’s AI-powered video creation and information platforms. According to VANK, these systems have occasionally failed to provide accurate historical context or have blocked the generation of content related to the islands, which are currently administered by South Korea but claimed by Japan. The organization asserts that such algorithmic biases reflect a lack of cultural and historical nuance in large language models (LLMs) developed outside of the regional context.
“We are witnessing a situation where AI platforms are effectively treating undisputed sovereign territory as a ‘disputed region’ based on biased training data,” VANK stated in its recent appeal. The group emphasizes that because AI models are trained on vast datasets primarily sourced from the global internet, they often inherit the geopolitical perspectives embedded in those sources. For South Korea, this creates an urgent need for “sovereign AI”—systems built on domestic datasets that accurately reflect the nation’s historical records and legal standing.
The Risks of Algorithmic Bias in Geopolitical Records
The core of the issue lies in how generative AI interprets and summarizes information from diverse, often conflicting, online sources. When an AI platform encounters a topic with historical friction, it often defaults to a neutral or “both sides” stance, which critics argue can effectively erase historical facts. In the case of Dokdo, South Korea maintains that the islands are historically and legally part of its territory, a position supported by various government records and historical maps. However, when AI platforms ingest data from international sources that may label the region as disputed, the resulting output can perpetuate that narrative to global users.
This phenomenon is not unique to Dokdo. Tech analysts have long noted that AI models often reflect the “majority” or “dominant” viewpoints present in their training data. As the National Institute of Standards and Technology (NIST) has highlighted in its AI Risk Management Framework, data quality and the mitigation of bias are critical to ensuring that AI systems remain reliable and trustworthy. For organizations like VANK, the failure to prioritize regional historical accuracy is a direct threat to the digital representation of national identity.
Advocating for Sovereign AI Infrastructure
VANK’s call to action is part of a broader discourse in South Korea regarding the necessity of building domestic AI infrastructure. The South Korean government has previously announced plans to foster a robust AI ecosystem, aiming to reduce dependence on foreign technology giants. According to the Ministry of Science and ICT, the government has been actively investing in the development of Korean-language-specialized LLMs to ensure that domestic culture, language, and historical context are accurately represented in the digital sphere.
Building a sovereign AI does not mean isolating a nation from global technology; rather, it involves creating models that are trained on high-quality, verified, and locally curated datasets. This approach ensures that when users query sensitive topics, the AI provides responses grounded in the verified history of the region. Experts suggest that as AI becomes the primary interface through which the world consumes information, the ability to curate this data will become a fundamental aspect of digital sovereignty.
The Path Forward: Accountability and Transparency
As the debate continues, the focus remains on whether global AI providers will implement more transparent and localized moderation policies. Many users and advocacy groups are pushing for a “fact-check” mechanism within AI platforms, where historical claims are cross-referenced with official, peer-reviewed, or government-validated databases before being presented as objective fact.
For South Korean stakeholders, the goal is to ensure that international AI companies acknowledge the historical record regarding the Dokdo islands. The Ministry of Foreign Affairs of the Republic of Korea maintains extensive documentation outlining the historical and legal basis for its sovereignty over the territory, which it encourages international bodies and tech organizations to consult. The next phase of this advocacy will likely involve direct engagement with major tech firms to address how their AI models process, categorize, and report on regional historical disputes.
The conversation regarding AI sovereignty is expected to gain further traction during the next scheduled session of the National Assembly of the Republic of Korea, where lawmakers continue to discuss digital policy and the protection of national interests in the era of generative AI. We will continue to monitor developments regarding how international AI companies adjust their content moderation policies in response to these concerns. Please share your thoughts on the role of AI in shaping historical narratives in the comments below.