This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
NHK terminated its AI-powered multilingual subtitle service after the system, powered by Google Translate, incorrectly translated 'Senkaku Islands' as 'Diaoyu Islands' in Chinese subtitles during a news broadcast. The error, which occurred during coverage of Japan-U.S. summit talks, was deemed inappropriate by the broadcaster. NHK has since suspended the service in all nine languages to address the issue of unstable translations.
NHK announced the termination of its AI-powered multilingual subtitle service after an automatic translation error rendered "Senkaku Islands" as "Diaoyu Islands," the Chinese designation. The mistake, discovered on February 10th, 2025 during a news segment, led NHK to deem the service inappropriate. The AI-based subtitles, powered by Google Translate, had been in use since 2020.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
Also classified as
The translation error resulted in the broadcast of incorrect and politically sensitive terminology ('Diaoyu Islands' instead of 'Senkaku Islands') to viewers.
The incident was caused by a lack of robustness in the AI translation system, which failed to reliably translate sensitive geopolitical terms, leading to incorrect outputs.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The translation error was an unexpected outcome produced by the AI system after it was deployed for live broadcasting.
NHK terminated its AI-powered multilingual subtitle service after Google Translate incorrectly translated the Japanese-controlled Senkaku Islands using the Chinese name Diaoyu Islands during a live broadcast. The incident highlights the risks of using fully autonomous translation tools for politically sensitive national broadcasts without real-time human oversight.
The AI system (Google Translate API) was used as a tool to perform a specific, well-defined function of translating English text into other languages in real-time based on direct inputs.
Minimal or No Risk: The system is a standard machine translation tool (Google Translate API) used for generating subtitles, which poses low risk to safety or fundamental rights under the EU AI Act definitions.
The AI system's sole purpose in this context was to translate English subtitles into multiple other languages for international viewers.
The reports from multiple reputable sources (NHK, Yomiuri, Yahoo Japan) consistently describe the incident, the specific translation error, the AI technology used (Google Translate API), and the subsequent action taken by NHK to terminate the service. There are no major contradictions.
None.
Technology
Unstable AI Translation Engine
Google AI translation API translated Senkaku to Chinese Diaoyu.
Lack of Custom Translation System
NHK relied on a generic third-party API instead of a specialized system.
No Post-Processing Term Filter
The system lacked a blocklist or translation override for sensitive terms.
Data Inputs
Biased Translation Training Data
The underlying AI model was trained on data associating Senkaku with Diaoyu.
Ambiguous Source Audio Input
Real-time speech-to-text or English subtitle inputs fed directly to AI.
Human Factors
Delayed Human Detection
An employee noticed the translation error only after it was broadcasted.
Over-reliance on Automated Systems
Staff trusted the AI to handle sensitive news without real-time oversight.
Process and Methods
Real-Time Direct Streaming
Subtitles were displayed live without a delay mechanism for review.
Inadequate Quality Control
No pre-broadcast validation process for automated translations.
Reliance on Simple Disclaimers
Using 'translation may not be perfect' warning instead of active control.
Management
Premature Service Deployment
Management deployed and expanded the service without resolving safety risks.
Lack of Risk Assessment
Management failed to assess geopolitical risks of automated translation.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
