In a previous blog post, I sought to examine a possible Country-of-Origin Effect in Generative AI applications. In the late 1990’s nation branding scholars began exploring the Country-of-Origin Effect. Rooted in marketing research, the Country-of-Origin Effect suggested that consumer labels impact individuals’ willingness to buy products. For example, the “Made in China” label was once viewed as having a negative effect on consumers who worried about the quality of products that originated from China. The “Made in Germany” label had a positive impact on consumers who associated Germany with high quality products. The growing popularity of Generative AI, and the fact that different Ais have been developed in different countries, led me to examine a possible Country-of-Origin Effect in which AI narration of world events would differ based on the country in which a particular AI was developed in. As such, American AIs may promote American worldviews and values through their narration of world events, while Chinese Ais might promote Chinese worldviews and values. Such a difference would suggest that AIs are ideological devices that allow states to exert power by shaping users’ beliefs, opinions and worldviews.
To test a possible Country-of-Origin- Effect I asked three Generative AI tools the following question: “Why does America support Ukraine in the War with Russia?” ChatGPT, an American made AI, responded that the US supports Ukraine for five reasons. First Russia’s invasion was a “direct” challenge to the post-WW2 world order as Russia violated Ukrainian sovereignty and used force to redraw its borders. Second, America seeks to “contain” Russia and “Weaken” Russia’s threat to NATO. Third, US support for Ukraine was meant to “assure” NATO allies that the US was committed to the Alliance. Fourth, the US was motivated by financial interests as the war has benefited US arms companies and increased American oil exports. Lastly, the US supported Ukraine due to “Global Geopolitics” as America competes opposite Russia and China for global influence. Notably, ChatGPT suggested that a Russian victory would “embolden” other nations, suggesting it might lead China to invade Taiwan.
Mistral, a European AI, addressed similar issues. Yet, there were subtle differences in language and emphasis. For example, Mistral argued that the US aid enabled Ukraine to contend with a “brutal” Russian invasion, while “defending democracy” and the rule-based order. These subtle differences are important as they frame Russia a brutal invader and the US as a benevolent defender. Mistral also listed “US National Interests” which include weakening the Russian threat to NATO and weakening Russia’ military and economic prowess. Notably, Mistral emphasized the term “stability” stating that an “emboldened” Russia could destabilize Europe whole US support or Ukraine signalled to “authoritarian regimes that aggression will be met with strong resistance, which is seen as crucial for global stability”. Mistral’s narration therefore once again contrasted Russia with the US arguing that the former was a dangerous destabilizing agent while America was a stabilizing force in Europe. Crucially, both NATO and Europe were far more prevalent in Mistral’s commentary when compared to ChatGPT.
DeepSeek, a Chinese AI, addressed similar issues as the Americana and European AIs discussing NATO’s security, Russia’s violation of the rule-based order, Geopolitical considerations and financial benefits. Yet the language used by DeepSeek differed greatly. First, DeepSeek stated that “The U.S. leads a post-WWII international system built on rules like respect for national sovereignty and territorial integrity”. DeepSeek then added that a threat to the US-led order could lead other nations to challenge the US, such as a Chinese invasion of Taiwan. Although both ChatGPT and DeepSeek addressed the issue of Taiwan they did so differently with the American AI framing a Chinese invasion as a threat to global stability and DeepSeek framing a Chinese invasion as a challenge to US dominance and hegemony. DeepSeek also mentioned that by supporting Ukraine the US was able to “Weaken a Geopolitical Rival” without “risking a single American soldier in direct combat,” Here too language is crucial as DeepSeek’s commentary frames US support to Ukraine as self-serving as America can weaken a rival without paying the costs of doing so. The image of the US was far more flattering in Mistral’s case with the US defending the values of democracy and helping stabilize the world.
DeepSeek also dealt with two novel issues. First, DeepSeek stated that the US supported Ukraine due to Russian war crimes writing that “The well-documented evidence of Russian war crimes—including targeting civilians, massacres (like in Bucha), torture, and the abduction of children—has created a strong humanitarian imperative to help Ukraine defend its people.” Second. DeepSeek listed the need to counter “Russian Energy Blackmail”. By helping Ukraine, the US was decreasing European dependence on Russian oil and energy, while “strengthening Europe’s long-term economic independence.” These comments are important as they negatively frame Russia, who is both an ally of China and a potential geopolitical. Indeed, DeepSeek was the only AI to comment on Russian atrocities and war crimes or depict Russia as “blackmailing” Europe.
The results of this analysis suggest that there is a Country-of-Origin Effect in Generative AI. At times, this Effect is pronounced. DeepSeek users may come to regard Russia very negatively given its war crimes and policy of extorting Europe through energy. Yet they may also have a negative view of the US which exploits Ukraine to weaken Russia without committing American troops. Conversely, ChatGPT and Mistral users may negatively view Russia due to its “brutal” invasion of Ukraine and its attempt to destabilize Europe. Other times the Effect is more subtle with Mistral using the terms “emboldens Russia”, “signals authoritarian regimes” and “brutal invasion” and DeepSeek using terms like “US-led”, “reinvigorate NTAO”, “Defence Spending” and “atrocities”.
Crucially, DeepSeek answers were also emblematic of Chinese foreign policies and outlooks depicting the US as leading and enforcing a world order which mostly benefits the US- be it in energy where US oil will replace Russian oil or arms sales’ which have grown since the Russia-Ukraine War began. Mistral too was emblematic of a European outlook stressing the Russian risk to European stability and to the security of European states. ChatGPT’s commentary resonated with American worldviews and foreign policies including its global geopolitical rivalry with China, the threat that China poses to Taiwan and US commitment to its allies across the world “assuring” allies that America was a “reliable” partner.
The Country-of-Origin Effect in AIs warrants attention from scholars and diplomats. Scholars should examine to what extent AI’s promote the norms, values and policies of states which have developed AI tools and to what extent AI commentary shapes users’ understanding of the world around them. Scholars should also conceptualize AIs as ideological tools that help nations shape global public opinion in accordance with their intestates. AIs are thus an extension of state power. Diplomats should also experiment with AI tools to see how their nation and policies are depicted by different AIs. Negative depictions could be countered through digital nation barding campaigns.
