Sovereignty has never meant an unlimited right to impose harm beyond national borders. AI and other global risks may require that principle to be extended to new forms of transnational danger.
The first part of this piece argued climate change, AI, unregulated online medicine and pandemics expose regulatory problems that traditional arms-control arrangements cannot simply solve. The harms are diffuse, incentives to defect can be strong, and in some cases the border itself no longer contains the activity being regulated.
This raises a deeper question: if sovereignty is part of the obstacle, is the Peace of Westphalia the problem? Not quite – and understanding why points towards what would actually have to change.
The Peace of Westphalia was two treaties, signed at Münster and Osnabrück in 1648, that ended the Thirty Years’ War and the Eighty Years’ War. They rearranged the map of central Europe and settled, imperfectly, a set of religious and dynastic disputes among German princes. What they did not do is declare a general doctrine of sovereign equality and non-intervention among states. That doctrine, so-called Westphalian sovereignty, was substantially assembled afterwards, by later jurists looking for a tidy origin for the state system they already had. Andreas Osiander’s influential study called it, bluntly, a myth. Stephen Krasner went further, arguing that even sovereignty norms genuinely established in the modern era have never been consistently honoured, routinely overridden through convention, coercion and plain imposition whenever it suited a state to do so. Sovereignty, on Krasner’s account, has always been “organised hypocrisy”, a rule invoked more often than followed.
The legal architecture a reformer needs to confront is therefore not sitting in an archive in Münster. It lies in the modern international order built around sovereign equality, territorial integrity, non-intervention and limits on the use of force. The United Nations Charter’s provisions are central, but even here some care is needed. Article 2(7) limits intervention by the United Nations in matters essentially within a state’s domestic jurisdiction; the broader principle that states must not coercively intervene in one another’s affairs also exists in customary international law. Sovereignty is thus not an individual rule, principle or instrument that can simply be rewritten. It is an architecture of Charter provisions, customary law and institutions.
The real question, then, is not whether to tear up a 1648 peace treaty, but whether an international order organised principally around sovereign states can adequately govern harms that originate within one jurisdiction but impose risks far beyond it.
Responsibility to Protect (R2P) offers one precedent, albeit a deliberately narrow one. After Rwanda and Bosnia exposed what an absolutist conception of non-interference could permit inside a state’s borders, the 2005 UN World Summit affirmed that every state has a responsibility to protect its population from genocide, war crimes, ethnic cleansing and crimes against humanity. Where a state manifestly fails, the international community accepted a responsibility to respond through the United Nations, using peaceful means and, where necessary, collective action under the Charter.
R2P therefore did not abolish sovereignty or create a general right of intervention. It reframed sovereignty as carrying responsibilities as well as rights. Its record since has been varied. The NATO intervention in Libya in 2011 was subsequently seen by Russia, China and many other states as having stretched a civilian-protection mandate into regime change, damaging trust in the doctrine. On Syria, proposed action ran into Russian and Chinese vetoes while several hundred thousand people died. R2P shows that sovereignty can be framed as responsibility, but also how, when major powers disagree, that framing can be significantly limited.
A second, more successful answer is the European Union – a sharing of sovereignty. Having suffered that two-part war, its member states voluntarily transferred specific, named powers, over trade, consumer and competition policy, and increasingly climate and digital regulation, to a supranational body they created. This is the closest working model of what a genuine climate or AI floor would require. The EU is an institution able to set an enforceable carbon border price or a compute threshold. It has real jurisdiction over at least some of the commons problem. It took decades and a self-selected club of wealthy, adjoining democracies to develop. Canada and maybe other nations becoming ‘associate members’ of the EU perhaps offers the possibility of the Union’s regulatory principles operating more widely on the globe.
It is conceivable that other regional institutions could, in time, move toward the EU model to deal with commons problems the same way. But climate and AI governance do not have that kind of time. They are moving too fast and their consequences are too grave. What they need is a way to bring the United States, China and Russia into the same room now.
A third development comes from international environmental law. It recognises the Common Concern of Humankind: the idea that some problems are of such shared importance that they cannot sensibly be treated as exclusively domestic. The Paris Agreement’s own preamble expressly recognises climate change as a ‘common concern of humankind’. But common concern does not itself create an international regulator or a free-standing power to override national sovereignty. Its significance is normative: it recognises the legitimate interest of the international community in harms whose consequences are shared.
There is an even firmer precedent. International environmental law has long balanced sovereign rights with a principle of prevention or ‘no harm’. Principle 2 of the Rio Declaration recognises a state’s sovereign right to exploit its own resources, but couples that right with responsibility to ensure that activities within its jurisdiction or control do not damage the environment of other states or areas beyond national jurisdiction. The International Court of Justice has recognised this underlying obligation as part of international environmental law. That does not make environmental law automatically applicable to artificial intelligence. But it does establish an important proposition: sovereignty has never meant an unlimited legal entitlement to exercise domestic freedom while exporting serious harm beyond one’s borders.
So to deal with AI, we do not need to dust off the 1648 treaties. The environmental no-harm principle already deals with externalised risk; the challenge is whether its underlying logic can be adapted to risks that travel through digital networks, global supply chains or biological systems rather than the climate and waterways.
We might call that broader principle a responsibility not to externalise: no sovereign or corporate actor should enjoy an unlimited entitlement to capture the benefits of an activity while imposing catastrophic or long-tail risks on people beyond its jurisdiction. The principle would not erase sovereignty. It would qualify its exercise where the consequences are foreseeably and substantially borne by others.
A normative declaration would be a plausible way to state a principle like that without waiting for 60 ratifications or a unanimous Security Council resolution. The 1948 Universal Declaration of Human Rights was not itself a binding treaty and was adopted with eight abstentions. But it did reshape, over the following decades, what conduct by a sovereign state the world will tolerate without comment. A similar kind of instrument has been proposed by one of us for a related threshold: an expanded rights declaration, issued with something like the UDHR’s own authority, setting out how humanity should engage with increasingly autonomous AI systems, before any threshold is crossed rather than after. Responsibility to Protect is a caution against expecting too much of such a document alone; Syria is the lesson in what happens when a great power declines to be bound. But the UDHR’s own history is the correction. Even without enforcement powers, a declaration can gradually shift the burden of justification from those trying to prevent harm to those claiming the right to cause it.
International practice already shows that sovereignty can coexist with responsibilities to populations, can be pooled voluntarily between states, and can be constrained by obligations to prevent transboundary harm. None of these precedents provide a ready-made regime for artificial intelligence, pandemics or digitally enabled cross-border harms. But together they challenge the idea that sovereignty can mean an unlimited freedom to externalise consequences.
A world that was always ecologically interconnected now faces hazards that are technologically global as well. Developing a further principle for catastrophic commons and border-erosion risks would therefore not abandon sovereignty. Indeed, it could strengthen it by continuing a long process of defining the responsibilities that come with it.
If AI carries even a fraction of the risks its own developers now warn of, that task may not remain theoretical for long.
Read Part One of this two-part series:
Robin Brown
Robin Brown is Deputy Chair of Fairer Future. He has advised Australian and overseas governments, businesses and NGOs on consumer protection, competition policy and regulatory accountability. Formerly head of the Consumers’ Federation of Australia, he helped secure the landmark court ruling that enabled bans on second‑hand tobacco smoke and spearheaded creation of both the Consumers’ Health Forum and the Australian Communications Consumer Action Network.
Gregory is a D’harawal man from NSW. He worked at SES levels in the Departments of Foreign Affairs and Trade, Environment, Climate Change, and Indigenous Affairs. He represented Australia as an Ambassador and High Commissioner in West Africa, and negotiated for Australia in the United Nations Climate Change negotiations. Gregory was Australia’s first-appointed Threatened Species Commissioner. He led development and implementation of Australia’s first Threatened Species Strategy and first Threatened Species Prospectus. In the NGO sector, he led, grew and reformed a national Indigenous charity.
Gregory speaks French and Mandarin. He has an Honours Degree in Economics and a Masters in Foreign Affairs and Trade.