This research investigates how large language models from different geopolitical origins (US vs China) exhibit biases when responding to politically sensitive topics.

Key findings:

• LLMs reflect the geopolitical perspectives of their countries of origin
• These AI systems can potentially shape political narratives and influence public opinion
• The biases appear in how models handle internationally contentious topics
• Security implications arise when these systems are deployed globally

For security professionals, this research highlights the critical need to assess AI systems for hidden geopolitical biases that could impact international relations, information integrity, and national security decision-making.

Original paper: Echoes of Power: Investigating Geopolitical Bias in US and China Large Language Models