Summary: This post argues that Artificial General Intelligence (AGI) threatens both liberal democracy and rule-based international order through a parallel mechanism. Domestically, if AGI makes human labor economically unnecessary, it removes the structural incentive for inclusive democratic institutions—workers lose leverage when their contribution is no longer essential. Internationally, if AGI gives one nation overwhelming productivity advantages, it erodes other countries' comparative advantages, reducing the benefits of trade and weakening incentives to maintain a rule-based world order. The post draws historical parallels to early 20th century concerns about capital concentration, distinguishes between "maritime" (trade-dependent) and "continental" (autarkic) power strategies, and discusses what middle powers like the EU might do to remain relevant. The core insight is that both democracy and international cooperation rest on mutual economic dependence—and AGI could eliminate both dependencies simultaneously.
Read this if you're interested in: AGI's geopolitical implications, how economic structures shape political systems, the future of liberal democracy, or strategic options for countries that won't lead in AGI development.
Epistemic status: fairly speculative and likely incomplete or inaccurate, though with a lot of interesting links.
Introduction
The Effective Altruism community has long acknowledged the risks of AGI, especially those related to the loss of control, for instance via gradual disempowerment). Less attention has received the issue of stable totalitarianism, AI-powered totalitarianism that could more easily enforce large-scale surveillance; and extreme power concentration, where a handful of companies or countries might have a much larger degree of power and might challenge the concept of liberal democracies.
This post examines the last of these risks—extreme power concentration—but not through the lens of a coup or sudden takeover. Instead, I focus on structural forces that create incentives for liberal democracy and rule-based international order, and how AGI might erode both simultaneously through a parallel mechanism.
Here's my core argument: Both liberal democracy and rule-based international order rest on structural incentives created by mutual dependence. Internally, the need for human labor creates incentives for inclusive institutions. Externally, the benefits of trade based on comparative advantage create incentives for rules-based cooperation. AGI threatens to weaken both dependencies simultaneously—reducing the value of human labor domestically and comparative advantage internationally. This parallel erosion could undermine the foundations of the current democratic and rule-based world order.
I'll first examine how liberal democracies survived early concerns about capital concentration because labor remained economically essential, and why AGI presents a qualitatively different challenge. Then I'll analyze how AGI could shift major powers from trade-oriented "maritime" strategies toward autarkic "continental" strategies, weakening rule-based order. Next, I'll discuss bottlenecks that might slow these dynamics and provide leverage for maintaining cooperation. Finally, I'll explore what AI middle powers—Europe in particular—might do to remain relevant.
A quick disclaimer: I am not a historian nor an economist, so there may be important gaps in my arguments. This is an exposition of my current understanding, offered to invite discussion and correction.
Leverage without labour
In the late 19th and early 20th centuries, many intellectuals considered socialism or communism superior to capitalism. For observers at the time, it seemed plausible that the capitalist model emerging from the Industrial Revolution would hand control of nascent democracies to a small elite with ownership of the means of production. Historical examples like the East India Company—which effectively functioned as a state in the territories it controlled—suggested that extreme capital concentration could indeed override formal political structures.
These concerns proved exaggerated. While communism failed to deliver prosperity and often devolved into authoritarianism, liberal democracies survived and thrived. Today's reality is more nuanced: billionaires may have large political influence, but the welfare state has expanded significantly, and most citizens in developed democracies enjoy unprecedented material prosperity and political rights.
A key structural factor could explain democracy's resilience: labor remained essential to production. I have sometimes read the argument that this is due to labour remaining a strong complement to capital, and in fact, labour remaining a roughly stable (~60%) fraction of GDP (Labor share of gross domestic product (GDP) - Our World in Data). More importantly, the complementarity between labor and capital meant that governments needed worker cooperation for economic growth and military capacity, workers retained leverage through their ability to withhold labor (strikes) and votes, and inclusive institutions outperformed extractive ones because investing in human capital—education, healthcare, infrastructure—generated returns through increased productivity.
AGI represents a qualitatively different challenge from previous automation waves. Previous technological advances, from the steam engine to computers, increased the productivity of human labor rather than replacing it. Workers could be retrained for new roles. The worry is that if AGI renders the marginal value of human work close to 0, then some or most of the incentives for liberal democracy and inclusive institutions could disappear.
This is not simply another shift in which tasks humans perform—it is a fundamental change in whether human labor remains economically necessary. For some time, institutional inertia might keep things going, but there is a distinct chance that there might be a significant erosion of democratic institutions in such a world over the long term (Anton Leicht: AI and Jobs).
If a small group can achieve prosperity without broad-based human contribution, the long-term equilibrium may drift toward more extractive institutions. And even if we ultimately establish good institutions in this new equilibrium, I believe people will be extremely confused about which actions we should take during the transition, just as I believe many early communist intellectuals had good intentions. This is thus a precautionary tale about taking strong actions early on, when the best path forward is not clear yet.
The “maritime” world order
Just as liberal democracy has proven to be the most promising government structure for growth and prosperity, the rule-based world order born since the Second World War has also shown to have many advantages. Here too, we find a structural foundation: this order enables countries to participate in trade, which creates mutual gains through comparative advantage.
David Ricardo's insight was that even if one country could produce every good more efficiently than another, both benefit from specialization and trade. Each country focuses on what it produces relatively most efficiently, enabling all participants to grow richer. This creates a powerful incentive to maintain the rules and institutions that facilitate trade—international law, freedom of navigation, dispute resolution mechanisms, and so forth. Trade, in turn, strongly depends on this idea of relative advantage: even if one country might on its own be better at producing any single product or service, by specializing and devoting resources where they are more productive, they leave space to other countries to grow their own relative advantage. And thus, by trading and following rules, countries may grow richer and more prosperous (Sarah C. Paine / Noahpinion Interview).
Strategic thinkers distinguish between "maritime powers" that depend on trade and therefore invest in maintaining open sea lanes, international rules, and alliance networks, versus "continental powers" that prioritize territorial control and self-sufficiency. The post-WWII American order has been fundamentally maritime: the U.S. maintained global rules and alliances because it benefited from the resulting trade network.
But just as AGI threatens the value of human work, it could grant the U.S. such an extreme economic advantage that other nations see their relative advantage significantly eroded. The question becomes: why trade with Germany for precision manufacturing when AI systems can match or exceed that capability? Why maintain alliance commitments to secure access to Japanese technology when those capabilities can be indigenized?
If the U.S. no longer gains marginal utility from foreign specialized markets, the functional incentive to maintain a rule-based order weakens significantly. The U.S. shifts from a "Maritime Power" (invested in global rules) to a more autarkic "Continental Power" that views allies not as partners in a mutually beneficial system, but as potential liabilities or strategic buffer zones.
The shift would likely be toward a more transactional order rather than complete autarky. The U.S. would still need physical resources it lacks (rare earth minerals, certain energy sources), consumer markets for AI-enabled products and services, coalition partners to counterbalance rivals, and the prevention of adversary counter-blocs. However, these needs create weaker incentives for a rule-based order than for mutual comparative advantage. They lead to bilateral deals based on narrow interests rather than broad multilateral frameworks. Partners become valued not for shared governance principles but for specific resources or strategic positions they control.
Even with AGI, certain goods and services will remain scarce or expensive. Physical resources like energy, rare earth minerals, agricultural land, and water cannot be produced by intelligence alone. Regulatory and political approval processes resist automation. Human-centric services where human interaction is valued for its own sake may resist full automation. Manufacturing facilities, data centers, and energy infrastructure require physical presence—geography still matters.
Thus, it seems that at least for some time, countries that may get a grip on some part of the value chain might still be treated as allies by the U.S. This is to some extent what is already happening with the current U.S. administration, even if not caused by the development of AI: we seem to be transitioning from a world where alliances are based on values to one where alliances are based on just having something the other country needs—what Anton Leicht calls "Pax Silica" (Anton Leicht, Forging a Pax Silica), a play on Pax Americana but based on silicon/computing rather than maritime power.
Anton Leicht believes this has a good side, making alliances less dependent on the sympathies of the administration, but I fear it is less stable than one may think. Even if most U.S. allies currently have some leverage on parts of the AI value chain, it seems likely the U.S. government will seek to indigenize as much value from the value chain as it can (US Congress, CHIPS Act 2022). And even without government support, there will be private attempts to challenge that status (e.g. SemiAnalysis: How to Kill 2 Monopolies with 1 Tool).
Further, changes in which countries dominate those technologies typically happen in the time scales of 1 or 2 decades, whereas those time scales are too short to cultivate stable alliances, which post WW2 have lasted significantly longer. Additionally, AI might dramatically reduce the coordination costs and inefficiencies that currently limit how quickly large organisations can expand into new markets.
A normative aside on transactional alliances
In the next paragraphs, I will discuss some thoughts on what AI middle powers might do about this. However, first, I provide a very personal opinion on how this transactional approach to a world order makes the U.S. look in other liberal democracies. In short, quite bad (Pew Research Center, Views of the United States). Instead of being considered primus inter pares, the U.S. starts being viewed as one of the bullies (e.g. Russia, China) that goes around imposing their conditions for their own exclusive benefit. In fact, it is hardly news that sometimes the current U.S. administration treats the Vladimir Putin government better than its allies.
This is not to say the U.S. has to become the world police or solve others' problems at their expense. For instance, as a European, I think Europe should be responsible for its security with little help. However, I think it would be wrong to assume the U.S. can choose a transactional relationship with its allies and they will just remain so because they are democracies. There is a historical precedent of a democracy (India) allying with autocratic countries (the Soviet Union) over the U.S. during the Cold War (Sarah Paine — The war for India); and there is some incentive for Europe and China to partially ally and isolate Russia (Sarah Paine – How Russia sabotaged China's rise).
It is for this reason that I wish liberal democratic values to remain an important part of how the U.S. develops alliances, not just a pure transactional approach. Instead, I argue that respect for individual freedom and fundamental rights—values many would call American values—should be the main reason to treat other countries as partners and allies.
A Roadmap for the AI Middle Powers
In any case, beyond what the U.S. might do, it is worth considering what middle powers might do to remain relevant. Moreover, I believe the EU might hold a fairly unique responsibility as a democratic "super-state" large enough to provide redundancy as a large democratic world power (CFG, Europe and the geopolitics of AGI).
There are quite a few things the EU should do.
First, the economic foundations: It is important to have a competitive energy market and deep capital markets (EU-Inc, Draghi Report on EU Competitiveness), and deepen its economic ties with the rest of the democratic world and specially emerging powers like India (Noah Smith, Europe is under siege). Europe can also leverage its diverse ecosystems to quickly experiment and find what policies work best, and then propagate them quickly (Noah Smith, Four thoughts on eurosclerosis).
Second, technology policy: The EU should also have a pro-growth stance on technology and AI, facilitating applications of AI, while remaining committed to safeguarding systemic risks (Luis Garicano, The EU and the not-so-simple macroeconomics of AI, Luis Garicano, The constitution of innovation). Some argue that aiming to commoditise the model layer, ensuring the portability of data and ownership of agents, and creating a vibrant application ecosystem might not only help prevent gradual disempowerment (Ryan Greenblatt, The best approaches for mitigating "the intelligence curse") but also help maintain geopolitical power (Luis Garicano, The smart second mover). Unfortunately, I am less optimistic about the latter advantage. Technology companies today remain somewhat constrained to their core competencies by workforce coordination challenges. AI agents could remove this bottleneck, enabling rapid invasion of adjacent markets.
Third, value chain positioning: The AI middle powers should aim to keep hold of the value chain as much as possible. Private and specialised sources of data might, if properly protected, provide some durable hold beyond the ever-changing technological edge (IFP: Unlocking a Million Times More Data). Additionally, robotics might be an area that is not yet as capital-intensive and scale-dependent, and Europe holds important know-how here.
Fourth, electrification: It might be beneficial for AI middle powers to specialise in the electric stack technologies (Not Boring: The Electric Slide). This would provide some highly needed independence from China in key areas, and complement the U.S. focus on software and AI. After all, two are the key inputs to production: energy to produce stuff, and intelligence to direct that energy. This interest in electrification may capitalize Europe's interest in green tech, not just for climate reasons but for long term productivity growth too.
Finally, public goods provision: The EU might be able to provide public goods that the U.S., with its constant discussion of racing with China, might not want or be able to provide (CFG: Building CERN for AI). This includes research in AI safety or on best practices on AI, perhaps allowing it to shape global standards.
There are many reasons to be pessimistic about the European Union: it is slow, it typically overregulates and has little chance to become a competitive player in the development of AGI. On the other hand, probably in a biased way, I think Europe and the European Union structurally have more built-in infrastructure for democracy than any other region. Not only are the majority of states in the region small and highly interdependent, but the European Union also has instruments to limit the authoritarian tendencies some of the national governments may exhibit as a consequence of their ideological pursuits (e.g. Hungary). The European Union is often plagued by the need for consensus between member states (Pieter Garicano, Policies without politics), but that same lack of speed characterises democracy vs autocracy, and allows democratic countries to slowly course-correct when they make mistakes. Some in the American right believe the E.U. is a bureaucratic instrument of the left (Heritage foundation), or the only place where communism succeeded (Noah Smith, Europe is under siege). This is wrong. It is often a technocratic point of view – instead of ideological –, or a strong consensus on the matter that dictates what policies are implemented in the E.U. Meanwhile, politics is often dominated by national politics, whose governments still hold most of the political power and arguably are the main reason for the lack of speed in implementing much-needed reforms (Draghi Report on EU Competitiveness). In any case, Europeans feel quite positive towards the E.U. (Eurobarometer 2025 Winter survey). For all the reasons above, I believe the E.U. may have an important role to play in how liberal democracy survives in the upcoming age of AGI.
Conclusion
AGI poses parallel threats to the structural foundations of both domestic liberal democracy and international rule-based order. Internally, it risks making human labor economically unnecessary, removing a key incentive for inclusive institutions. Externally, it risks making trade less valuable by eroding comparative advantages, removing a key incentive for rules-based cooperation. These are not certainties, but structural pressures that will shape the post-AGI world.
The risks are greatest if we approach the transition with excessive confidence in our understanding of the right path forward. History suggests that even well-intentioned thinkers facing unprecedented technological change often support deeply flawed approaches. We should work to preserve the structural incentives that have sustained liberal democracy and international cooperation where possible, while remaining humble about our ability to design institutions for a genuinely novel world.
Much of the responsibility for navigating this transition lies with major powers, particularly the U.S. and potentially China. However, middle powers—especially large democratic blocs like the E.U.—have roles to play in maintaining redundancy in the global system, controlling key bottlenecks, and providing public goods. The window for establishing these positions may be measured in years or decades, but it will not remain open indefinitely.
The stakes are not merely national prosperity, but the persistence of the liberal democratic model that has, despite its flaws, enabled unprecedented flourishing over the past century. That model rests on foundations that AGI will test as profoundly as any force in modern history.
Summary: This post argues that Artificial General Intelligence (AGI) threatens both liberal democracy and rule-based international order through a parallel mechanism. Domestically, if AGI makes human labor economically unnecessary, it removes the structural incentive for inclusive democratic institutions—workers lose leverage when their contribution is no longer essential. Internationally, if AGI gives one nation overwhelming productivity advantages, it erodes other countries' comparative advantages, reducing the benefits of trade and weakening incentives to maintain a rule-based world order. The post draws historical parallels to early 20th century concerns about capital concentration, distinguishes between "maritime" (trade-dependent) and "continental" (autarkic) power strategies, and discusses what middle powers like the EU might do to remain relevant. The core insight is that both democracy and international cooperation rest on mutual economic dependence—and AGI could eliminate both dependencies simultaneously.
Read this if you're interested in: AGI's geopolitical implications, how economic structures shape political systems, the future of liberal democracy, or strategic options for countries that won't lead in AGI development.
Epistemic status: fairly speculative and likely incomplete or inaccurate, though with a lot of interesting links.
Introduction
The Effective Altruism community has long acknowledged the risks of AGI, especially those related to the loss of control, for instance via gradual disempowerment). Less attention has received the issue of stable totalitarianism, AI-powered totalitarianism that could more easily enforce large-scale surveillance; and extreme power concentration, where a handful of companies or countries might have a much larger degree of power and might challenge the concept of liberal democracies.
This post examines the last of these risks—extreme power concentration—but not through the lens of a coup or sudden takeover. Instead, I focus on structural forces that create incentives for liberal democracy and rule-based international order, and how AGI might erode both simultaneously through a parallel mechanism.
Here's my core argument: Both liberal democracy and rule-based international order rest on structural incentives created by mutual dependence. Internally, the need for human labor creates incentives for inclusive institutions. Externally, the benefits of trade based on comparative advantage create incentives for rules-based cooperation. AGI threatens to weaken both dependencies simultaneously—reducing the value of human labor domestically and comparative advantage internationally. This parallel erosion could undermine the foundations of the current democratic and rule-based world order.
I'll first examine how liberal democracies survived early concerns about capital concentration because labor remained economically essential, and why AGI presents a qualitatively different challenge. Then I'll analyze how AGI could shift major powers from trade-oriented "maritime" strategies toward autarkic "continental" strategies, weakening rule-based order. Next, I'll discuss bottlenecks that might slow these dynamics and provide leverage for maintaining cooperation. Finally, I'll explore what AI middle powers—Europe in particular—might do to remain relevant.
A quick disclaimer: I am not a historian nor an economist, so there may be important gaps in my arguments. This is an exposition of my current understanding, offered to invite discussion and correction.
Leverage without labour
In the late 19th and early 20th centuries, many intellectuals considered socialism or communism superior to capitalism. For observers at the time, it seemed plausible that the capitalist model emerging from the Industrial Revolution would hand control of nascent democracies to a small elite with ownership of the means of production. Historical examples like the East India Company—which effectively functioned as a state in the territories it controlled—suggested that extreme capital concentration could indeed override formal political structures.
These concerns proved exaggerated. While communism failed to deliver prosperity and often devolved into authoritarianism, liberal democracies survived and thrived. Today's reality is more nuanced: billionaires may have large political influence, but the welfare state has expanded significantly, and most citizens in developed democracies enjoy unprecedented material prosperity and political rights.
A key structural factor could explain democracy's resilience: labor remained essential to production. I have sometimes read the argument that this is due to labour remaining a strong complement to capital, and in fact, labour remaining a roughly stable (~60%) fraction of GDP (Labor share of gross domestic product (GDP) - Our World in Data). More importantly, the complementarity between labor and capital meant that governments needed worker cooperation for economic growth and military capacity, workers retained leverage through their ability to withhold labor (strikes) and votes, and inclusive institutions outperformed extractive ones because investing in human capital—education, healthcare, infrastructure—generated returns through increased productivity.
AGI represents a qualitatively different challenge from previous automation waves. Previous technological advances, from the steam engine to computers, increased the productivity of human labor rather than replacing it. Workers could be retrained for new roles. The worry is that if AGI renders the marginal value of human work close to 0, then some or most of the incentives for liberal democracy and inclusive institutions could disappear.
This is not simply another shift in which tasks humans perform—it is a fundamental change in whether human labor remains economically necessary. For some time, institutional inertia might keep things going, but there is a distinct chance that there might be a significant erosion of democratic institutions in such a world over the long term (Anton Leicht: AI and Jobs).
If a small group can achieve prosperity without broad-based human contribution, the long-term equilibrium may drift toward more extractive institutions. And even if we ultimately establish good institutions in this new equilibrium, I believe people will be extremely confused about which actions we should take during the transition, just as I believe many early communist intellectuals had good intentions. This is thus a precautionary tale about taking strong actions early on, when the best path forward is not clear yet.
The “maritime” world order
Just as liberal democracy has proven to be the most promising government structure for growth and prosperity, the rule-based world order born since the Second World War has also shown to have many advantages. Here too, we find a structural foundation: this order enables countries to participate in trade, which creates mutual gains through comparative advantage.
David Ricardo's insight was that even if one country could produce every good more efficiently than another, both benefit from specialization and trade. Each country focuses on what it produces relatively most efficiently, enabling all participants to grow richer. This creates a powerful incentive to maintain the rules and institutions that facilitate trade—international law, freedom of navigation, dispute resolution mechanisms, and so forth. Trade, in turn, strongly depends on this idea of relative advantage: even if one country might on its own be better at producing any single product or service, by specializing and devoting resources where they are more productive, they leave space to other countries to grow their own relative advantage. And thus, by trading and following rules, countries may grow richer and more prosperous (Sarah C. Paine / Noahpinion Interview).
Strategic thinkers distinguish between "maritime powers" that depend on trade and therefore invest in maintaining open sea lanes, international rules, and alliance networks, versus "continental powers" that prioritize territorial control and self-sufficiency. The post-WWII American order has been fundamentally maritime: the U.S. maintained global rules and alliances because it benefited from the resulting trade network.
But just as AGI threatens the value of human work, it could grant the U.S. such an extreme economic advantage that other nations see their relative advantage significantly eroded. The question becomes: why trade with Germany for precision manufacturing when AI systems can match or exceed that capability? Why maintain alliance commitments to secure access to Japanese technology when those capabilities can be indigenized?
If the U.S. no longer gains marginal utility from foreign specialized markets, the functional incentive to maintain a rule-based order weakens significantly. The U.S. shifts from a "Maritime Power" (invested in global rules) to a more autarkic "Continental Power" that views allies not as partners in a mutually beneficial system, but as potential liabilities or strategic buffer zones.
The shift would likely be toward a more transactional order rather than complete autarky. The U.S. would still need physical resources it lacks (rare earth minerals, certain energy sources), consumer markets for AI-enabled products and services, coalition partners to counterbalance rivals, and the prevention of adversary counter-blocs. However, these needs create weaker incentives for a rule-based order than for mutual comparative advantage. They lead to bilateral deals based on narrow interests rather than broad multilateral frameworks. Partners become valued not for shared governance principles but for specific resources or strategic positions they control.
Pax silica
The limits to the threat posed in the previous section are what economists call Baumol's cost disease (Pieter Garicano, Learning to love Baumol, J.Z. Mazlish: AK or just ok? AI and economic growth; Epoch AI, AI and explosive growth redux). As some parts of the economy see their productivity grow rapidly, other parts become the bottlenecks.
Even with AGI, certain goods and services will remain scarce or expensive. Physical resources like energy, rare earth minerals, agricultural land, and water cannot be produced by intelligence alone. Regulatory and political approval processes resist automation. Human-centric services where human interaction is valued for its own sake may resist full automation. Manufacturing facilities, data centers, and energy infrastructure require physical presence—geography still matters.
Thus, it seems that at least for some time, countries that may get a grip on some part of the value chain might still be treated as allies by the U.S. This is to some extent what is already happening with the current U.S. administration, even if not caused by the development of AI: we seem to be transitioning from a world where alliances are based on values to one where alliances are based on just having something the other country needs—what Anton Leicht calls "Pax Silica" (Anton Leicht, Forging a Pax Silica), a play on Pax Americana but based on silicon/computing rather than maritime power.
Anton Leicht believes this has a good side, making alliances less dependent on the sympathies of the administration, but I fear it is less stable than one may think. Even if most U.S. allies currently have some leverage on parts of the AI value chain, it seems likely the U.S. government will seek to indigenize as much value from the value chain as it can (US Congress, CHIPS Act 2022). And even without government support, there will be private attempts to challenge that status (e.g. SemiAnalysis: How to Kill 2 Monopolies with 1 Tool).
Further, changes in which countries dominate those technologies typically happen in the time scales of 1 or 2 decades, whereas those time scales are too short to cultivate stable alliances, which post WW2 have lasted significantly longer. Additionally, AI might dramatically reduce the coordination costs and inefficiencies that currently limit how quickly large organisations can expand into new markets.
A normative aside on transactional alliances
In the next paragraphs, I will discuss some thoughts on what AI middle powers might do about this. However, first, I provide a very personal opinion on how this transactional approach to a world order makes the U.S. look in other liberal democracies. In short, quite bad (Pew Research Center, Views of the United States). Instead of being considered primus inter pares, the U.S. starts being viewed as one of the bullies (e.g. Russia, China) that goes around imposing their conditions for their own exclusive benefit. In fact, it is hardly news that sometimes the current U.S. administration treats the Vladimir Putin government better than its allies.
This is not to say the U.S. has to become the world police or solve others' problems at their expense. For instance, as a European, I think Europe should be responsible for its security with little help. However, I think it would be wrong to assume the U.S. can choose a transactional relationship with its allies and they will just remain so because they are democracies. There is a historical precedent of a democracy (India) allying with autocratic countries (the Soviet Union) over the U.S. during the Cold War (Sarah Paine — The war for India); and there is some incentive for Europe and China to partially ally and isolate Russia (Sarah Paine – How Russia sabotaged China's rise).
It is for this reason that I wish liberal democratic values to remain an important part of how the U.S. develops alliances, not just a pure transactional approach. Instead, I argue that respect for individual freedom and fundamental rights—values many would call American values—should be the main reason to treat other countries as partners and allies.
A Roadmap for the AI Middle Powers
In any case, beyond what the U.S. might do, it is worth considering what middle powers might do to remain relevant. Moreover, I believe the EU might hold a fairly unique responsibility as a democratic "super-state" large enough to provide redundancy as a large democratic world power (CFG, Europe and the geopolitics of AGI).
There are quite a few things the EU should do.
First, the economic foundations: It is important to have a competitive energy market and deep capital markets (EU-Inc, Draghi Report on EU Competitiveness), and deepen its economic ties with the rest of the democratic world and specially emerging powers like India (Noah Smith, Europe is under siege). Europe can also leverage its diverse ecosystems to quickly experiment and find what policies work best, and then propagate them quickly (Noah Smith, Four thoughts on eurosclerosis).
Second, technology policy: The EU should also have a pro-growth stance on technology and AI, facilitating applications of AI, while remaining committed to safeguarding systemic risks (Luis Garicano, The EU and the not-so-simple macroeconomics of AI, Luis Garicano, The constitution of innovation). Some argue that aiming to commoditise the model layer, ensuring the portability of data and ownership of agents, and creating a vibrant application ecosystem might not only help prevent gradual disempowerment (Ryan Greenblatt, The best approaches for mitigating "the intelligence curse") but also help maintain geopolitical power (Luis Garicano, The smart second mover). Unfortunately, I am less optimistic about the latter advantage. Technology companies today remain somewhat constrained to their core competencies by workforce coordination challenges. AI agents could remove this bottleneck, enabling rapid invasion of adjacent markets.
Third, value chain positioning: The AI middle powers should aim to keep hold of the value chain as much as possible. Private and specialised sources of data might, if properly protected, provide some durable hold beyond the ever-changing technological edge (IFP: Unlocking a Million Times More Data). Additionally, robotics might be an area that is not yet as capital-intensive and scale-dependent, and Europe holds important know-how here.
Fourth, electrification: It might be beneficial for AI middle powers to specialise in the electric stack technologies (Not Boring: The Electric Slide). This would provide some highly needed independence from China in key areas, and complement the U.S. focus on software and AI. After all, two are the key inputs to production: energy to produce stuff, and intelligence to direct that energy. This interest in electrification may capitalize Europe's interest in green tech, not just for climate reasons but for long term productivity growth too.
Finally, public goods provision: The EU might be able to provide public goods that the U.S., with its constant discussion of racing with China, might not want or be able to provide (CFG: Building CERN for AI). This includes research in AI safety or on best practices on AI, perhaps allowing it to shape global standards.
There are many reasons to be pessimistic about the European Union: it is slow, it typically overregulates and has little chance to become a competitive player in the development of AGI. On the other hand, probably in a biased way, I think Europe and the European Union structurally have more built-in infrastructure for democracy than any other region. Not only are the majority of states in the region small and highly interdependent, but the European Union also has instruments to limit the authoritarian tendencies some of the national governments may exhibit as a consequence of their ideological pursuits (e.g. Hungary). The European Union is often plagued by the need for consensus between member states (Pieter Garicano, Policies without politics), but that same lack of speed characterises democracy vs autocracy, and allows democratic countries to slowly course-correct when they make mistakes. Some in the American right believe the E.U. is a bureaucratic instrument of the left (Heritage foundation), or the only place where communism succeeded (Noah Smith, Europe is under siege). This is wrong. It is often a technocratic point of view – instead of ideological –, or a strong consensus on the matter that dictates what policies are implemented in the E.U. Meanwhile, politics is often dominated by national politics, whose governments still hold most of the political power and arguably are the main reason for the lack of speed in implementing much-needed reforms (Draghi Report on EU Competitiveness). In any case, Europeans feel quite positive towards the E.U. (Eurobarometer 2025 Winter survey). For all the reasons above, I believe the E.U. may have an important role to play in how liberal democracy survives in the upcoming age of AGI.
Conclusion
AGI poses parallel threats to the structural foundations of both domestic liberal democracy and international rule-based order. Internally, it risks making human labor economically unnecessary, removing a key incentive for inclusive institutions. Externally, it risks making trade less valuable by eroding comparative advantages, removing a key incentive for rules-based cooperation. These are not certainties, but structural pressures that will shape the post-AGI world.
The risks are greatest if we approach the transition with excessive confidence in our understanding of the right path forward. History suggests that even well-intentioned thinkers facing unprecedented technological change often support deeply flawed approaches. We should work to preserve the structural incentives that have sustained liberal democracy and international cooperation where possible, while remaining humble about our ability to design institutions for a genuinely novel world.
Much of the responsibility for navigating this transition lies with major powers, particularly the U.S. and potentially China. However, middle powers—especially large democratic blocs like the E.U.—have roles to play in maintaining redundancy in the global system, controlling key bottlenecks, and providing public goods. The window for establishing these positions may be measured in years or decades, but it will not remain open indefinitely.
The stakes are not merely national prosperity, but the persistence of the liberal democratic model that has, despite its flaws, enabled unprecedented flourishing over the past century. That model rests on foundations that AGI will test as profoundly as any force in modern history.