Leveraging AI in Public Diplomacy

The rapid development of AI tools has caused a frenzy in foreign ministries (MFAs) as diplomats across the world are trying to identify the risks and benefits brought about by artificial intelligence tools such as ChatGPT, Mistral, Claude, Gemini and DeepSeek. Diplomats’ attempts to grapple with the professional and societal ramifications of AI has taken different forms in different MFAs with some hoping to establish in-house AI tools and others dedicating staffers to mapping how AIs could be leveraged to obtain foreign policy goals.

Thus far, diplomats seem to be focusing on two main areas. The first is the automation of routine diplomatic functions. This very notion demonstrates how societal discourses may shape diplomats’ attitudes towards digital technologies. As newspapers and industry leaders all predict that AIs will soon replace entire segments of the labor force thanks to automation, diplomats also come to view the benefits of AI through the narrow lens of automation, be it through consular "bots" that could replace consular officers or the automatic formulation of press releases, speeches and social media content. However, some MFAs have already discovered that while Generative AIs excel at producing texts, these are usually very formulaic and tend to follow a small number of templates. Generative AI texts are thus generic and fail to elicit interest or emotions. One diplomat who read an AI generated address to the UN Human Rights Council stated, “no one would remember this address five minutes after it was given and no journalist would cover it.”

In-house AI tools are viewed by diplomats as a potential “game changer” especially if these tools could be used to analyze the vast amounts of digital data that MFAs produce on a daily basis including emails, reports, ministerial briefings, analyses of ties between states, reports on state visits, crisis management and negotiation. The allure of AI is that it may facilitate data-informed policy making in diplomacy. Yet this entails a dangerous assumption: that the future will mirror the past. Indeed, an in-house AI tool may be used to analyze previous rounds of negotiations between states while helping diplomats formulate successful negotiation tactics. Similarly, in-house AI tools may be used to review press statements by foreign countries and identify terminology that is indicative of crisis escalation. Yet the difficulty of crisis management lies in the fact that crises are often novel and are dissimilar to past ones. Russia’s invasion of Ukraine differed from the stealthy annexation of Crimea, while the COVID-19 pandemic differed from Ebola outbreaks.

"The allure of AI is that it may facilitate data-informed policy making in diplomacy. Yet this entails a dangerous assumption: that the future will mirror the past."

However, there is one area where Generative AI may be of use to MFAs and that is in gauging global public opinion. AI tools such as ChatGPT are trained on vast amounts of readily available digital information. This includes websites, social media posts, comments on news sites, free sites such as Wikipedia and blogs. As such, Generative AIs like ChatGPT may be conceptualized as aggregators of global public opinion. If asked to list ten good things about France, ChatGPT’s answer would be based on commonly held perceptions and opinions about France. If asked to list ten bad things about Nigeria, ChatGPT’s answer would essentially be a summary of commonly held beliefs about Nigeria. Lastly, if asked why the U.S. supports Ukraine, or why Germany has enacted strict migration policies, ChatGPT’s answers would again be an aggregation of popular notions and beliefs.

This suggests that MFAs may use Generative AI as a tool for gauging global public opinion regarding a nation’s image, its reputation, its policies and the role it plays on the global stage. This insight may then be used by diplomats when narrating state action or crafting campaigns to enhance a state’s image. MFAs may also use AI outputs to challenge misconceptions about nations and regions. For instance, African nations may dedicate digital efforts to countering misperceptions found across AIs which depict Africa as inherently corrupt, insecure and plagued by social unrest. It is thus in the reals of public diplomacy and nation branding that AIs may prove especially beneficial to diplomats.

Additionally, as an aggregator of digital information, Generative AIs reproduce biases found online. For example, AIs often describe Global South countries as being very violent while failing to mention violence in Global North countries such as the U.S. Similarly, Global South countries are labeled as corrupt while AIs fail to note rampant corruption found in Global North countries such France and the UK. Social unrest is also now pervasive across the Global North but is rarely acknowledged by Generative AIs. Such biased depictions of world regions stem from the data used to train AIs including Wikipedia and blog posts which themselves are rife with stereotypes and misconceptions.

Notably, Generative AIs were not designed to be aggregators of opinions and beliefs, and they are mostly trained on English-language content which means that they aggregate information written by English speakers. Yet, the depiction of states, regions and policies in and across Generative AIs may serve as an initial data set used by diplomats to gauge public opinion and enhance their communication strategies and their ability to “sell” their foreign policies. In this sense, AI may be a “game changer” in public diplomacy.