The Pope, the President, and the G7: what the AI governance power contest means for middle states

By Eleonora Lamm

Also available in Spanish

During the last month, three significant documents on AI governance materialised: the G7 Ministerial Declaration on Digital and Technology, adopted in Paris on May 29 under the French presidency; the Executive Order on Promoting Innovation and Security in Advanced Artificial Intelligence, signed by President Trump on June 2; and, by far the most unexpected contribution, the Magnifica Humanitas, an encyclical on AI published by Pope Leo XIV on May 15.

Rarely have three documents so distinct in origin and authority converged on the same issue in such a short time. That the Vatican, the White House, and the G7 are all discussing AI simultaneously is no coincidence: it is the clearest sign that the governance of this technology has become one of the major issues of our time.

In my more than 20 years working at the intersection of law, ethics and technology, I have come to see that the documents policymakers produce often reveal more than they intend, and in many cases they surface aspirations about what kind of governance is actually possible.

Taken together, these three texts reveal the tensions, aspirations, and power realities shaping the AI governance landscape in 2026. They enable us to explore the reality of international AI governance and, in particular, to consider the  implications for so-called ‘middle countries’.

Three documents. Three claims on AI power

The Executive Order opens with the declaration that the United States leads in AI ‘because we refuse to stifle…innovation with overly burdensome regulation’. This unambiguous ideological stance positions deregulation as the engine of American technological dominance. In keeping with this view, the White House envisions a ‘voluntary framework’ for frontier model access as the appropriate governance instrument. Voluntary as it might be, this represents a step forward – there appears to be a growing interest in controlling risks, but control, after all, is also power.

The G7, meanwhile, calls for AI public procurement guidelines while also celebrating broader private-sector participation. It elevates the Hiroshima AI Process Reporting Framework to the status of a ‘key multi-stakeholder platform’, and by doing this it proposes a different vision: that the legitimacy of AI governance requires broad and diverse participation, not just bilateral agreements between states and large corporations.

The encyclical seems to grasp where the debate is heading. While it argues that it is necessary to adopt ‘adequate regulatory tools capable of upholding justice and curbing the distorting effects of technological power’, it clarifies that ‘the issue is not limited to regulation’. What the Pope identifies – and what my own experience confirms – is that we need governance in addition to regulation. And governance encompasses norms, institutions, incentives, multi-stakeholder processes, and the social, cultural, and economic frameworks within which technology is built and used.

Leo XIV argues that a world in which powerful private actors exceed the governance capacity of states – compounded by the retreat of multilateral institutions – creates precisely the conditions for AI to become an instrument of domination rather than an engine of human development. The widespread reliance, by both the G7 and the White House, on voluntary industry participation is symptomatic of a new reality – a reality in which an inherent imbalance has potentially immense and dangerous implications.

From recipients to architects: the strategic imperative for middle countries

Against this backdrop, countries that are neither technological powers nor passive recipients of digital globalisation – what we at Oxford Insights call ‘transitioning’ or ‘middle’ countries – face a distinct challenge: the strategic construction of their own governance capacity and, more specifically, their autonomy in the 21st century.

Data sovereignty and computing capacity are, in 2026, structural conditions of political autonomy. A country that cannot process its own public health data, or lacks the capacity to audit algorithms affecting judicial decisions, has neither sovereignty over AI nor meaningful governance of it. This does not mean technological autarky, which would be unviable. It means building domestic institutional capacity: oversight agencies with genuine technical competence; legal frameworks that establish transparency and auditability requirements; and regional digital infrastructure agreements that reduce dependence on single providers and help rebalance the distribution of power.

In real terms, supporting middle countries means delivering something concrete: helping governments develop technical expertise; identifying regulatory models that can be adapted to local institutional realities; building regional coalitions that give smaller actors collective leverage; and documenting what actually works – not in theory, but in practice, in specific contexts, with specific constraints.

I increasingly believe in the strategic role that middle states can – and should – play as active builders of their own governance. After all, AI governance is not merely a technical problem that can be solved with a technical solution. It is, at its core, an ethical and political problem that demands ethical and political solutions. 

Success will ultimately depend on the quality of the institutions we build, the breadth of the voices we include, and the political will to treat AI for what it is: not a neutral tool, but a civilisational choice – one that middle countries have both the right and the responsibility to help shape.

Grid image by Coronel G on Unsplash