The US decision last month to ban international access to Anthropic’s latest model revealed a growing gap between the pace of technology and the ability of policy to keep up. Then, later in June, the fact that the G-7 leaders gave away much of their summit shows that they see the stakes. But also has been revealed the absence of an actionable agenda.
In this matter as in others, geopolitics limits what is possible. The imperative to stay ahead constrains what countries can do domestically and what they can do together. But a measure like the Anthropic ban, which in its transparency as an export control did not distinguish America’s partners from others, risks an unexpected result: technological fragmentation that weakens the West’s collective level, which is important for its strategic advantage.
The US decision last month to ban international access to Anthropic’s latest model revealed a growing gap between the pace of technology and the ability of policy to keep up. Then, later in June, the fact that the G-7 leaders gave away much of their summit shows that they see the stakes. But also has been revealed the absence of an actionable agenda.
In this matter as in others, geopolitics limits what is possible. The imperative to stay ahead constrains what countries can do domestically and what they can do together. But a measure like the Anthropic ban, which in its transparency as an export control did not distinguish America’s partners from others, risks an unexpected result: technological fragmentation that weakens the West’s collective level, which is important for its strategic advantage.
The accumulation of freedom in Europe fully detached from the United States is not attainable or desirable for Europe or the United States. Frontier’s ability to scale America is beyond the power of Europe on any convenient schedule, and the capital of Europe is well placed within the ranks in the important places of dispatch and planning, where it can gain agency and develop its strategic advantage. The United States, meanwhile, should not have a strategic interest in pushing its allies toward such self-sufficiency. But the ban, which undermines allies’ assurance of consistent access to US technology, threatens that very much. Building a global architecture to prevent these outcomes is not only urgent but strategically important.
What is needed is a pre-agreed playbook and a pre-prepared tool that can be intentionally deployed without collateral damage to ensure safe release and consistent access. The open nature of AI, and open source in particular, makes this really difficult. As a challenge to collective action based on national security and linked to geopolitical competition, it is up to states to determine a predictable basis for security delivery. The good news is that the safety instructions—and the risk mitigation that the latest models have revealed—are clearer than the current confusion suggests.
Of England The Bletchley Park summit in 2023, which brought together the United States, China, and nearly 30 other countries, solid a basic international architectural framework on AI security: international regulations covering the most serious cross-border risks, including bio- and cyber security; and collaboration between partners and companies on testing and reliable information sharing, supported by adequate government capacity. In a different way, this is reflected in the recent proposals from the leaders of the frontier labs themselves. If the system hadn’t been rolled back last year by executive order, the Trump administration would have had more than a deployment ban when borderline models showed the dangers the system predicted.
But while the security challenges are high, the economic impact is low, widespread, and highly political. That makes the economy tougher—and arguably more productive. The the question for many people it is not security in the abstract but as AI adds their functions or removes them automatically. The politics there are local. But because many countries are, to varying degrees, adopters rather than creators of the technology, the drivers that determine how AI grows and spreads in the economy—on everything from access to capital—are global. AI security governance has a framework, however it is not complete. Economic governance does not.
There will be no Bretton Woods moment for AI. The governance of emerging technologies will change aggressively and rapidly. There are, however, parallels with precedents. The international nuclear regime, and the International Atomic Energy Agency in particular, is many times is mentioned. The parallels continue to this day, but show that cooperation on shared risk is possible even at the height of strategic competition. The Financial Stability Board (FSB)—which it started as a platform and has grown into something with real appeal—it may be the most underappreciated model. His lesson is that sustainable architecture is built through successive layers of collaboration, not a single act of design. Waiting for that clarity will mean missing time. Instead, governments should start with practical steps, accepting that the agenda will change with uncertainty.
The first of those practical steps is to establish a common understanding of what the future of AI in policy is. International AI Security Report that he went out the Bletchley summit provides a good basis for assessing technological progress. That needs to be complemented by better understanding in its deployment. That should not come from a single source (there is no monopoly of wisdom), but a system is needed—this could be the Organization for Economic Cooperation and Development—by recognizing which international policy directives apply in all situations and which depend on different future paths.
The second step is to build coordination facilities, starting from when interests meet and a model exists. The latest frontier models have revealed cyber-related financial stability risks. But there is a much wider range of risks of large-scale fragmentation – loss of asset prices, production shocks, market concentration – that can be systemic and merit the most prudent and counter-intuitive tools developed since the financial crisis. This would be a good place to start. The FSB is already for him.
The third is to deal with partner cooperation—and its imperative. If fragmentation is to be avoided, the priority is to put in place safeguards that ensure the certainty and consistency of access to hardware and software. While the US administration seeks to agree with border laboratories on a more sustainable procedure for the safe delivery of their samples, this aspect of the partners should not be ignored. Furthermore, it is an incomplete agenda to ensure that, as domestic AI regimes evolve, they do not differ in ways that create barriers between partner markets, and even more so, as AI spreads throughout the economy, to ensure that the rules written for a pre-AI world, on everything from cross-border services terms to reach standards. suitable for one AI.
The fourth is to combine this with strategic support for the global deployment of AI in emerging markets and developing countries. Affordable access will be important for the prosperity of the country and the contours of geographical and political competition. The US and its allies need to learn from technologies like 5G in this regard. The next generation of development policy should have this at its core, and it can only be achieved through cooperation between governments and companies that develop and distribute technology.
The fifth goes to the heart of international politics: how countries share in the revenue from AI as the benefits inevitably focus. Thoughts like that public equity shares in major AI companies on the list shows that the internal debate within the US is starting. Outside of the United States, the question is how the countries that help create value for these companies, not least through their data and customers, let alone any innovations produced there, can share in the revenue. This income will be important in managing the disruption of the transition. Although politically difficult, intellectual work inside and outside government about the answers needs to start now, and elected officials should initiate the debate.
None of these steps can develop a suitable speed without a machine to carry them. What is needed first is a technical layer that can work under politics and maintain cooperation, including the International Civil Aviation Organization, the International Maritime Organization, the International Telecommunication Union, and, to some extent, the World Trade Organization. On top of that must come the political architecture that must be governed. International regulations will require UN legitimacy despite its limitations.
Any management of potential risk will have to come from negotiations between the US and China. Partner collaboration will require something more focused. The G-7 plus partners are the best place to start, especially as a forum for the US and other nations to address the dynamics between them. But it will need to evolve into something more permanent and systematic—not a grand convention, but something that can sustain action over time. That will ultimately require action from the United States, but a set of others may fall. Britain holds the presidency of the G-20 next year; it should build on what it started at Bletchley Park and make this its top priority.
The ban on Anthropic’s latest model may take it away, but the gap it’s exposed won’t close on its own. A dysfunctional default, with unfavorable consequences for geographic competition and competitive markets, as well as safety and security, is already emerging to fill it. Creating an architecture to replace that is not the work of the future. It is for now.
