Ben Harack contributes to proposal for multinational AI project

DPIR’s Ben Harack – a DPhil in International Relations student - was part of the team that produced A Blueprint for Multinational Advanced AI Development last month, alongside the world's most cited living scientist Yoshua Bengio and a number of other key figures. 

Aspirationally, the memo could be important for countries like the UK who have significant AI capabilities but are nonetheless far behind the US and China today. 

We caught up with Ben to ask him more about the crucial memo and how he felt to be a part of such a significant and impactful project… 

What is the key message you were trying to get across in the paper? 

Mid-sized economies face significant challenges in the emerging age of AI, but they are very well-positioned to address them if they work together. 

Can you explain your involvement in the project? 

Charles Martinet led the production of the memo. I was one of the main people providing direct support and writing throughout 2025. 

Can you explain in a little more detail what the paper is about? 

In a nutshell, the paper is about why mid-sized economies should work together to forge their own path on frontier AI. The context for the paper is twofold:

1. The largest cutting-edge AI models (termed "frontier AI") have become crucial for economic and military competitiveness. Given the pace of progress and the abilities of current systems, it is reasonable to infer that AI will play an increasingly important role in national competitiveness in the future.
2. The US and China are currently far in the lead on the production of frontier AI. Alone, no mid-sized economy can feasibly compete with them in this domain. 

The key danger for mid-sized economies (and in fact all countries in the world) is that they may face an extremely undesirable choice: either 

• become dependent on frontier AI produced by one of the two leading states, or
• relinquish competitiveness by not integrating frontier AI. 

The first option erodes sovereignty by making the state dependent on another state for its competitiveness. The second option means that the state's economy and military will be weaker than its peers. 

The opportunity outlined in the paper is for a third option which addresses the challenge in a way that safeguards both sovereignty and competitiveness. Mid-sized economies are well-positioned to develop frontier AI together in a multinational project. While the resources needed to produce competitive frontier AI models are beyond the reach of mid-sized economies working alone, a group of these countries working together can compete robustly. In such a collaboration, these countries can draw upon strengths that are not widely known, such as the remarkable proportion (80+%) of top AI researchers that hail from these countries. 

Mid-sized economies would benefit from such collaboration in several ways. 

1. They would retain reliable access to frontier AI capabilities.
2. They would gain an important seat at the table in the international governance of frontier AI.
3. They would foster larger domestic AI ecosystems which provide robust competitive advantages even in futures where the multinational collaboration dissolves. 

Finally, such a multinational collaboration preserves options for these countries in the face of uncertainty. If AI proves to be deeply transformative, the collaborative project can safeguard the sovereignty and competitiveness of its members and broadly serve the interests of humanity. If AI proves to be only minimally transformative, the collaborative project will have fostered deeper and more effective investments in the domestic AI ecosystems of member states, which can in turn produce benefits for those states. 

Can you explain its relevance in today’s society? 

The age of AI came roaring into prominence with the launch of ChatGPT in late 2022. In less than three years, there are now billions of people using frontier models. An avalanche of investments from individuals, companies, and countries around the world indicates that there is broad belief that general-purpose AI will be broadly transformative and will provide durable advantages to those who produce it first. If the benefits were not believed to be both substantial and long-lasting, investment of the scale we see would not be occurring. 

Unless mid-sized economies take action, there is a high likelihood that the story of AI will be largely written by the US and China, who currently host the two largest AI development ecosystems (by a large margin). Mid-sized economies (and the rest of the world) face the prospect of being essentially sidelined in the competition between the two great powers. 

Taking the initiative in this relationship can serve not only domestic interests, but humanity as a whole. If the bipolar structure of the frontier AI economy continues to consolidate, we could face a new Cold War with myriad disturbing consequences for the world. If the middle powers work together, they can be a strong third voice in such a world. Even if we still end up facing a new Cold War, these countries acting together can provide a way for humanity to converge on an AI future that is broadly beneficial. While a bipolar world might have structural pressures pushing it toward militarization and even war, a world with a neutral, humanist, and collaborative third option is a world with a better chance of preventing war and unlocking a brighter future. 

How did you feel to be part of this? 

The research and writing process was daunting, but I'm now proud of what we produced. It was daunting because these are thorny and complicated issues with potentially dramatic consequences for our societies and perhaps even the entire future of humanity. Over many iterations, the analysis evolved into a shape that appears to hold up to serious scrutiny. While it's guaranteed to be imperfect, I nonetheless think that we produced something that is worth considering.