Why AI will Change the Core of Foreign Policymaking

Cornelius Adebahr explores the way AI may help the people, processes and policies that make up the foreign policy system.

Much of the discussion on how artificial intelligence (AI) will affect international affairs is conducted in broad strokes and grandiose terms. That AI-assisted weapon systems, whether as “killer robots” or “killer algorithms”, will make future wars unpredictable and unaccountable. That the technology itself represents an existential threat to humanity on par with nuclear annihilation. Or, simply, that whoever leads in AI will rule the world.

All of that is possible, of course, and international efforts to rein in the uncontrolled spread of AI are commendable. Yet, such scenarios must not blind politicians nor analysts to the more immediate changes the technology will bring to the way foreign policy is conducted. Because these will come for sure, and they are – for the moment, at least – still very much in control of ministries and chanceries.

Two things will force a fundamental rethink to the art of diplomacy, even before we consider a shifting international landscape: the technology behind AI, and the fact that it is mostly in the hands of private companies. The former may be less obvious than the latter, but no less profound.

As I’ve experienced myself when co-creating an online toolkit explaining “Democracy by Design” to employees of tech start-ups, the translation of political concepts into an application requires hard thinking about how to express what in which technical terms. Rather than explaining something from one human being to another (which is what diplomats often do), information, policies, and strategies have to be mediated by computerised processes. These, in turn, are programmed by people that are not at all in the foreign policy know but think in terms of actionable inputs and outputs.

So, in order to make use of the wonderful world of AI tools in front of them, foreign policymakers will soon have to clearly define certain inputs (like strategies, statistics, real-time data, meeting minutes, and the like) and how they should be processed in order to achieve any useful output. Or, in other words, the ‘secret sauce’ of diplomacy will have to be put into recipe form.

The second aspect, reliance on corporations, will also have profound implications. Today, much of foreign policymaking by officials is done with on-board means: personnel working from public buildings with proprietary equipment like encrypted communication tools (the frequent use of WhatsApp being the exception to the rule). Of course, governments have for decades bought computers and software for the worldwide use in ministries and embassies, but such helpful tools – like cars over carriages – merely facilitate the way diplomacy is conducted. They do not fundamentally alter it the way AI will.

In fact, for a sovereign to rely on specific custom-make AI models developed by companies requires a much greater amount of trust in, and control of, technology. In their own interests, governments will have to make sure that the products they purchase not only achieve the desired results but do so on sound legal and ethical grounds (think of copyright violations, in-built biases etc.). That’s obviously even more urgent in domestic policy areas concerning welfare benefits or legal rulings, where governments are likely to face lawsuits over the use of AI, but it is equally relevant in foreign policy.

Deconstructing strategies and policies into definable and executable parts and doing so in collaboration with tech experts from the private sector will thus fundamentally alter the way diplomacy is conducted already in the current decade. And while we’re at it, here’s how this could look like, in a good scenario. My ideas centre on 1) people, whose well-being should be the ultimate aim of foreign policy; 2) process, i.e., the way in which interests and preferences are matched with on-the-ground information to produce concrete instructions; and 3) policy, the product of the former two, which guides a government’s action towards its declared goals.

1. On people, I imagine a Facebook for Foreign Affairs, filled by diplomats with all their contacts and managed by an algorithm that updates it with information from the public domain, so that officials would know who to get in touch with, in which country or institution, on whatever subject matter, to establish direct communication.
2. On process, I can see data-driven horizon-scanning being widely employed, informed by actual events (such as political, social, kinetic, meteorological), searching for patterns and guided by predetermined parameters, with officials following up to verify and substantiate the tool’s findings.
3. On policy, I’d dream of a custom-built application based on large language models that allows the (dis)aggregation of policies into (or from) their integral parts (defined goals, operational measures, available resources and personnel, etc.), so that submissions or diplomatic cables or indeed strategies can be defined around certain, clearly recognisable factors.

There certainly are much better ideas out there, as foreign ministries, development agencies, intelligence bodies, and other government actors are trying to make use of the advent of AI. Still, after dreaming big and bold, things need to be boiled down to deliverables. Only then can AI contribute to improve foreign policymaking, and possibly also to avert the risk of extinction that the technology is said to embody.

Dr Cornelius Adebahr is a political analyst and consultant living in Berlin, Germany. His work focuses on European foreign policy issues, transatlantic relations, and Iran. Since the end of 2000, he has been the owner of Wirtschaft am Wasserturm – Political Consultancy, Project Development, and Training. In addition, he is a non-resident fellow at Carnegie Europe in Brussels and an associate fellow of the German Council on Foreign Relations (DGAP), as well as a member of the Team Europe of the European Commission.

Photo by Google DeepMind