People are increasingly turning to AI chatbots for news and explanations of world events. Yet the political answers they receive can vary with the language of the query. We show that the language used can markedly change how the same AI systems assess the war in Ukraine. We prompted GPT, Claude, and Gemini to evaluate twenty statements about the conflict in 112 languages, gathering 67,200 responses.\ \ The balance between Russia‑leaning and Ukraine‑leaning answers differs across languages. When responses are grouped by countries' official languages, a pattern emerges that mirrors global political divisions: languages with relatively more Russia‑leaning answers correspond to nations that hold more favorable public views of Russia, vote less in favor of Ukraine at the United Nations, and provide less aid to Ukraine. This broad pattern recurs across all three models and persists even after removing individual statement pairs.\ \ Our findings suggest a pathway whereby information warfare can shape the text used to train AI models, which in turn may propagate geopolitical biases.\ \ Review: The study provides a large‑scale, multilingual investigation of language‑dependent political bias in LLMs, highlighting a subtle channel through which bias can enter AI outputs and offering a foundation for future mitigation efforts.