Korea's Dominant Search Portal Told Its Own Users That Dokdo Is Japanese

What happened

In October 2025, Naver's AI search service labeled Dokdo as Japanese territory. Dokdo is a set of rocky islets in the sea between Korea and Japan that South Korea has administered since 1954 and considers an integral part of its sovereign territory. The mislabeling appeared in front of millions of Korean users on the country's most-used web portal, triggering immediate public outrage and prompting intervention at the level of the National Assembly.

The system behind the error is HyperCLOVA X, Naver's generative AI engine, which powers the search briefings the portal surfaces above traditional results. According to the incident record, the AI auto-ingested external public sources, including Japanese government materials, and summarized the Japanese territorial claim without flagging it as contested. Japan refers to the islets as Takeshima and has pressed its claim through official channels for decades. Naver's system apparently treated that official position as a factual summary rather than as one side of a live sovereignty dispute.

The political response was swift. Members of the National Assembly raised the issue publicly. A Korea University professor noted that while major Western AI systems such as ChatGPT frequently describe Dokdo's status as disputed, it was categorically different for Korea's own flagship portal to present the Japanese position as fact. Naver pulled the AI-generated briefings for Dokdo-related queries and pledged to review its content generation process to prevent a repeat.

The incident exposes two compounding failures. The first is technical: an AI summarization pipeline that treats all indexed public text as equivalent input, with no mechanism to weight or quarantine materials from parties to an active territorial conflict. The second is process-level: there was no review gate for topics where the stakes of a wrong answer are not factual but sovereign, and no domain expertise applied before the briefing reached users. Automated summarization that cannot distinguish a historical record from a contested government claim is not a neutral tool.

When Naver's AI produced its mislabeled briefing, there was no visible record of which specific sources drove the output, who reviewed the summary before it was served, or how long it had been live before complaints surfaced. That absence is the structural gap the incident makes concrete: a provable record of what a system pulled from, how it weighted those sources, and who cleared the result before publication would make errors like this auditable from the moment they appear, rather than reconstructable after the diplomatic damage is already done.

Reported impact

Affected parties

Not publicly disclosed

Harm type

Not publicly disclosed

Scale

Not publicly disclosed

Financial impact

Not publicly disclosed

Regulatory action

Not publicly disclosed

Classification

Organization

Not publicly disclosed

AI system

Not publicly disclosed

Industry

Not publicly disclosed

Country

Not publicly disclosed

Provider

Not publicly disclosed

Incident type

Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

• No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.