Globalization and the Problem of Exit

By Robert Ryan Jennings Published Sep 24, 2026 8:15PM 13 min read

society technology economics
globalization sovereignty ai data-sovereignty economics resource-based-economy
Globalization and the Problem of Exit

Part III of After Scarcity, Before Sovereignty: Sovereignty Is Having an Exit, What If Work Became Optional?, and Globalization and the Problem of Exit.

Globalization can produce substantial benefits. Trade lets regions specialize. Knowledge, capital, technology, and medicine can move farther and faster than they once did. Foreign investment can create infrastructure and productive capacity that might otherwise take decades to develop.

The problem starts when interdependence becomes dependency.

The twentieth-century version of globalization moved goods, factories, capital, and people across borders. The internet added something much stranger: identity, software, communication, operational control, and data could become global even when the people and institutions depending on them remained local.

AI may push that separation further. A person can interact with one interface while the work behind it crosses models, APIs, clouds, data centres, suppliers, and legal jurisdictions they never see.

We increasingly live in a world where the physical geography of society and the operational geography of society are not the same.

The internet crosses borders more easily than law does

Legal authority still attaches to jurisdictions, people, organizations, and property that exist somewhere. Courts have territorial jurisdiction, companies are incorporated somewhere, and physical assets generally have a location.

The mental model is roughly:

flowchart LR A[Person] --> B[Activity] B --> C[Jurisdiction] C --> D[Law]

The internet does not fit cleanly into that model.

A Canadian user can authenticate through an American identity provider, connect through a global CDN, use an application running in several regions, keep one database physically in Canada, create backups elsewhere, and call an AI model operated by another company in another country.

flowchart TD U[Canadian user] --> ID[Foreign identity provider] U --> APP[Application] APP --> CDN[Global network] APP --> DB[Canadian database] APP --> API[Foreign APIs] APP --> AI[AI model] DB --> BACKUP[Cross-border backups] ID --> CTRL[Foreign control plane] AI --> CTRL

So where did the activity occur?

Technically, perhaps in all of those places. Legally, that answer is much less useful.

The OECD’s work on cross-border data flows frames the same tension in policy terms: data flows underpin trade, logistics, communication, and digital services, while also raising questions about privacy, security, national security, regulatory reach, and government access.

Global infrastructure works remarkably well while the participants broadly trust one another. The difficult cases are the ones where that assumption fails.

Data residency is not data sovereignty

Data makes the mismatch unusually visible.

Keeping information physically inside a country’s borders can be useful. It does not answer every sovereignty question. A locally stored database may still depend on a foreign-owned cloud provider, foreign identity systems, remotely controlled encryption keys, overseas administrators, foreign software updates, or contracts subject to another country’s law.

The Government of Canada’s current Digital Sovereignty Framework extends this distinction beyond storage location. It addresses jurisdictional complexity, reliance on global suppliers, operational resilience, and institutional control as parts of digital sovereignty.

That gives us two different questions:

QuestionConcern
Where is the data physically stored?Residency
Who can ultimately compel, deny, alter, or administer access?Sovereignty

Data is also unlike a factory or a port. It can be copied without moving the original, replicated automatically, transformed into derived datasets, embedded in models, and accessed remotely across a border without any obvious physical transfer.

A foreign manufacturer may own a factory in Canada, but Canada at least has physical jurisdiction over the factory. A foreign digital platform can provide an essential Canadian service while much of its actual control plane remains elsewhere.

The dependency can look domestic until the moment sovereignty matters.

Corporations became the transport layer of globalization

Corporations are not inherently the problem with globalization. They are one of the main mechanisms through which its benefits happen.

Multinational companies are extremely good at moving capital, technology, labour, data, supply chains, standards, and organizational knowledge across borders. UNCTAD’s World Investment Report exists largely because foreign direct investment and multinational enterprises have become structural parts of the world economy.

flowchart LR C[Multinational corporation] C --> CAPITAL[Capital] C --> TECH[Technology] C --> DATA[Data] C --> SUPPLY[Supply chains] C --> LABOUR[Labour] C --> LAW[Legal structures]

The asymmetry is that corporate operations can span many jurisdictions while lawmaking, taxation, enforcement, and democratic accountability remain largely jurisdiction-specific.

A company can extract resources in one jurisdiction, manufacture in another, sell in a third, process data in a fourth, and recognize profits somewhere else. Pollution, waste, labour disruption, infrastructure costs, and ecosystem damage are much harder to move.

Foreign investment can genuinely create prosperity and still increase dependency. The risk does not require hidden intent: years of rational integration can erode fallback capacity until leaving becomes prohibitively expensive.

The problem of exit

A system should be judged not only by how efficiently it works during normal conditions, but by what happens when one participant needs to leave.

Imagine a country integrating deeply into a global service ecosystem for fifty years because the arrangement is cheaper and more capable than maintaining local alternatives. Then a government becomes hostile, a corporation fails, a trade relationship collapses, or an environmental externality becomes politically unacceptable.

Can the country still operate? Can its citizens export their data? Can critical software be moved? Can essential equipment be manufactured without permission from an overseas supplier? Can an AI service be replaced without rebuilding the systems around it?

flowchart TD G[Global integration] --> E[Efficiency + scale] G --> D[Dependency] D --> S[Geopolitical or institutional shock] S --> Q{Can we exit?} Q -->|Yes| F[Federation] Q -->|No| C[Captivity]

In software architecture, a dependency that cannot realistically be replaced is part of the trust boundary. Global systems deserve the same analysis.

This is the argument from Sovereignty Is Having an Exit applied at a larger scale. The answer is not necessarily less globalization. It is globalization with reversibility: portable data, interoperable standards, replaceable providers, local fallback capacity, and enough redundancy that leaving remains technically possible.

Spaceship Earth eventually wins the argument

There is a harder boundary beneath every political border: the planet itself.

Buckminster Fuller popularized the metaphor of Spaceship Earth: humanity shares one vessel with finite material stocks, bounded ecosystems, and no external place to send most of our waste.

At planetary scale, money is not the final constraint. It is an allocation system layered over physical constraints.

flowchart TD EARTH[Spaceship Earth] EARTH --> ENERGY[Energy] EARTH --> WATER[Fresh water] EARTH --> FOOD[Food + arable land] EARTH --> MINERALS[Minerals + materials] EARTH --> ECO[Ecosystems] EARTH --> WASTE[Waste absorption] EARTH --> COMPUTE[Compute + infrastructure]

The UN Environment Programme’s Global Resources Outlook 2024 reports that extraction of natural resources has tripled over roughly five decades and projects substantially greater extraction by 2060 without major changes. The planetary-boundaries research programme tries to quantify another part of the same problem: there are Earth-system processes within which human activity must remain if we want a stable operating environment.

A serious global order therefore cannot be based only on currencies, GDP, or political borders. At some layer it has to account for the physical system.

That does not imply one world government or one planning computer. It means that any system making claims about sustainable prosperity eventually needs a credible resource model: what exists, where it exists, how quickly it replenishes, what extraction damages, what reserves are necessary, and which forms of consumption are physically sustainable.

This is where the Venus Project’s resource-based economy is useful as a provocation even if its strongest conclusions are not accepted. Its important question is not whether money should disappear. It is whether the physical inventory underneath the monetary economy should become much more visible and explicit.

AI changes how much of the physical economy we can model

Comprehensive economic planning has long faced information and coordination problems. Prices remain powerful because they compress distributed information about scarcity and demand into signals that millions of actors can use without agreeing on a central plan.

AI does not magically eliminate that knowledge problem. It does change the observability of the physical economy.

Satellites, smart grids, logistics systems, inventories, weather models, agricultural sensors, shipping telemetry, factories, and supply-chain systems already produce extraordinary amounts of machine-readable information. AI can reconcile and forecast across those systems at a scale that earlier planners could not realistically attempt.

flowchart LR P[Physical world] --> S[Sensors + operational data] S --> R[Resource model] R --> A[AI forecasting + simulation] A --> H[Human policy constraints] H --> X[Production + allocation decisions] X --> P

The important architectural boundary is between observation, optimization, and governance.

flowchart LR O[Observation\nWhat exists?] --> P[Optimization\nWhat is possible?] P --> G[Governance\nWhat is legitimate?] AI1[AI + sensors] --> O AI2[AI models] --> P PEOPLE[Human institutions] --> G

AI may become much better than people at estimating crop yields, predicting energy demand, identifying waste, or simulating logistics. That does not give a model legitimate authority to decide which community loses water during a shortage.

Efficiency and legitimacy are different properties.

This connects directly to What If Work Became Optional?. If AI and robotics eventually provide much of the productive labour required by society, resource allocation becomes more explicit because employment no longer needs to remain the admission ticket to basic survival.

That would make the question less “who deserves income because they worked?” and more “what baseline can this physical system sustainably provide to every person, and which genuinely scarce goods still need markets, quotas, queues, lotteries, or other allocation mechanisms?”

Automation changes which differences become economic power

Automation does not make people equal, but it can change which differences matter economically. Machines can reduce the value of physical strength in some work, AI assistance can change the value of particular cognitive skills, and remote systems can reduce some geographic constraints.

The ILO’s 2025 work on generative AI and jobs is much more conservative than a post-work prediction: it finds broad exposure to GenAI but expects transformation to be more common than outright replacement under current conditions. The longer-term thought experiment still matters because automation can change which capabilities translate into income and bargaining power.

It can also create new concentrations of power around infrastructure:

Who owns the machines?
Who controls the models?
Who allocates compute?
Who controls energy?
Who defines the objectives?
Who can override the system?

AI can broaden access to capability while ownership of the underlying systems becomes more concentrated. That is why automation, globalization, and sovereignty cannot really be separated.

The internet may become infrastructure again

There is another possible change that makes jurisdiction stranger still. The human-facing internet may become smaller while the machine-to-machine internet becomes much larger.

Today the common model is still roughly:

flowchart LR H[Human] --> W[Website / app] --> C[Human-created content]

Increasingly, the human may deal with an AI interface while the underlying network becomes a substrate of APIs, agents, databases, sensors, and machines.

flowchart TD H[Human] --> AI[AI interface] AI --> API[APIs] AI --> DB[Databases] AI --> AGENT[Other agents] AI --> SENSOR[Sensors] AI --> MACHINE[Machines] AI --> WEB[Human web]

That does not mean the internet disappears. It may become less visible in the same way that most people stopped thinking about telephone switching equipment. Websites could matter less than machine-readable services. Search could matter less than synthesis. Content created mainly to attract human clicks or search traffic may lose much of its economic purpose.

Paradoxically, geography could become less visible to users while becoming more important strategically. A person may not know where an agent performed a task, but control of data centres, energy, fibre, satellites, identity systems, semiconductors, models, and cryptographic keys still belongs to somebody in some jurisdiction.

The interface becomes global while the dependency remains physical.

Global coordination without global captivity

Some problems really are planetary. Carbon, ocean pollution, biodiversity, pandemics, financial contagion, cyber incidents, and global supply chains do not respect political maps very well.

But the need for global coordination does not prove the need for a single global authority. A more resilient model looks like federation: common measurements, interoperable protocols, independently verifiable data, and local institutions that retain meaningful implementation choices and fallback capacity.

flowchart TD WORLD[Shared planetary constraints] WORLD --> A[Jurisdiction A] WORLD --> B[Jurisdiction B] WORLD --> C[Jurisdiction C] A <--> P[Shared protocols + measurements] B <--> P C <--> P A --> LA[Local capability] B --> LB[Local capability] C --> LC[Local capability]

Redundancy costs more than total integration, but it preserves fallback capacity when dependencies fail.

A useful global architecture would therefore preserve several properties that software engineers already recognize:

  • shared protocols rather than one implementation;
  • common measurements rather than one unquestionable authority;
  • interoperability rather than forced uniformity;
  • explicit trust boundaries;
  • independent verification of important state;
  • redundancy for critical capabilities;
  • graceful degradation when global dependencies disappear;
  • and a credible exit path.

The OECD’s “data free flow with trust” work is one existing example of this kind of tension. The objective is not simply to stop cross-border data movement, nor to pretend jurisdiction no longer matters, but to make interoperability possible while preserving protections and trust.

One instrument panel, many governments

There is an important distinction between agreeing about reality and agreeing about politics.

Countries can share measurements of an aquifer, fishery, mineral reserve, electricity grid, atmospheric concentration, or crop failure without agreeing on one government or one moral theory.

flowchart TD PANEL[Shared planetary instrument panel] PANEL --> FACTS[Common measurements] PANEL --> MODELS[Open models] PANEL --> RISKS[Shared risk forecasts] FACTS --> A[Jurisdiction A] FACTS --> B[Jurisdiction B] FACTS --> C[Jurisdiction C] A --> PA[Local policy] B --> PB[Local policy] C --> PC[Local policy]

Everyone can see the same fuel gauge and still disagree about where to drive.

That model also limits what AI is allowed to become. A planetary resource model could be enormously valuable if its inputs are inspectable, its assumptions are contestable, and independent institutions can reproduce important conclusions. It becomes much more dangerous if the same system owns the measurements, sets the objectives, allocates resources, operates the infrastructure, and decides whether an appeal is valid.

Greed, corruption, institutional capture, and simple incompetence do not vanish because a system is global or because an AI is involved. A design that works only when the people and models controlling it are benevolent is not a durable design.

The design question is what belongs at each level

The practical question is: at what level should each capability live, and can the lower level survive when the higher one fails?

Some resources and risks are planetary and demand global coordination. Some services become better and cheaper when they operate globally. Other capabilities are important enough that a society should retain a local fallback even when doing so appears inefficient during normal conditions.

The goal is not isolation, and it is not irreversible integration.

It is a world where cooperation is usually the best option, basic survival can be separated from compulsory employment, the physical constraints of our shared planet are visible, and communities retain enough sovereignty to leave institutions that stop serving them.

Earth may be a spaceship. That does not mean it needs one captain.

References and further reading