Taiwanese LLM Generates Controversial Political Content, Prompting Government Investigation

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

A Taiwanese large language model developed by Academia Sinica generated politically biased content, including identifying Xi Jinping as Taiwan's leader, sparking public controversy. The model was quickly taken offline, and the government launched an official investigation into the risks, data sources, and oversight of generative AI systems in Taiwan.[AI generated]

Why's our monitor labelling this an incident or hazard?

The article discusses concerns about the potential negative impacts of generative AI and LLMs in Traditional Chinese, including bias, misinformation, and misuse, but does not describe any realized harm or incident. Instead, it focuses on the challenges and the intention to investigate government policies and preparedness. This fits the definition of an AI Hazard, as it plausibly could lead to AI incidents if not properly managed, but no direct or indirect harm has yet occurred according to the article.[AI generated]
AI principles
AccountabilityFairnessTransparency & explainabilityRespect of human rightsDemocracy & human autonomySafety
Industries
Media, social platforms, and marketing
Affected stakeholders
General public
Harm types
ReputationalPublic interest
AI system task:
Content generation

Articles about this incident or hazard