AI Governance as Geopolitical Competition

Artificial intelligence is often discussed as a technological revolution.

Increasingly, it is also becoming a geopolitical one.

The competition is no longer limited to which company develops the most capable model.

Countries are competing over the infrastructure required to build those models, the chips required to train them, the data and talent required to improve them, the standards used to evaluate them and the rules that determine who can access them.

This changes the meaning of AI governance.

AI governance is no longer only about managing the consequences of technology.

It is increasingly connected to the distribution of technological power.

The central geopolitical question is therefore becoming:

Who has the capacity to build, control, evaluate and govern advanced AI?

The answer is not determined by models alone.

It depends on an entire technological ecosystem.

The AI Stack Is Becoming a Strategic Stack

A modern AI system depends on multiple layers.

At the hardware level there are advanced semiconductors, memory, networking equipment and specialised computing infrastructure.

Above that are cloud platforms and data centres.

Then come datasets, algorithms, foundation models and applications.

Around these technical layers sit standards, safety evaluations, regulatory institutions, intellectual property regimes, investment systems and skilled human capital.

Control over any one layer can influence the others.

A country may have strong AI researchers but limited compute.

Another may have large amounts of compute but weak frontier-model development.

Another may have advanced semiconductor manufacturing capabilities.

Another may possess regulatory influence without hosting the largest AI companies.

This means AI power is distributed across a complex global value chain.

Geopolitical competition is therefore increasingly about the whole stack.

The United States: AI Leadership and Strategic Infrastructure

The United States remains central to the global AI ecosystem because of its concentration of major AI companies, capital, cloud infrastructure, research institutions and advanced computing capacity.

But US AI policy increasingly treats AI capability as connected to national security and international competition.

The 2025 US AI Action Plan explicitly connects AI innovation, infrastructure, international diplomacy and security. It includes policies aimed at expanding American AI infrastructure and promoting US AI technology internationally.

The plan also describes advanced AI compute as strategically important and calls for stronger enforcement of controls intended to prevent advanced computing resources from reaching countries of concern.

This is an important development.

AI governance and technology-export policy are becoming increasingly interconnected.

A semiconductor export restriction can affect AI capability.

A cloud-access rule can affect model development.

A restriction on advanced computing can influence the strategic balance between countries.

Technology policy is therefore becoming part of AI governance.

China: AI Capability and Strategic Autonomy

China has developed a different relationship between AI development, national strategy and governance.

Its approach connects AI development with industrial policy, cybersecurity, data governance, standards and national technological capability.

At the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance in Shanghai, Chinese authorities explicitly framed AI governance as an international issue and called for stronger international cooperation, greater participation by developing countries and attention to fragmentation in global AI governance.

The same statement also called for AI agents to operate within clearly defined decision-making authority and behavioural boundaries, with mechanisms for tracing behaviour and identifying risks.

This is significant for two reasons.

First, China is increasingly presenting itself not only as an AI developer but also as a participant in the construction of international AI governance.

Second, the governance debate is becoming connected to competing ideas about how global AI rules should be organised.

The geopolitical issue is therefore not simply whether China develops advanced AI.

It is also whether Chinese institutions, standards and governance concepts become part of the international AI architecture.

Europe: Regulation Meets Technological Sovereignty

The European Union represents another important trajectory.

Europe initially became particularly influential through regulation, especially its risk-based AI governance model.

But regulation alone does not provide technological autonomy.

In 2026, the European Commission proposed a broader European Technological Sovereignty Package covering semiconductors, cloud, AI and open source. The package includes proposals for a Chips Act 2.0 and a Cloud and AI Development Act, alongside an EU Open Source Strategy.

This represents an important evolution.

Europe’s AI strategy is increasingly connecting:

regulation,

infrastructure,

compute,

semiconductors,

cloud,

open-source technology,

energy,

and industrial policy.

The EU is also expanding AI computing infrastructure through AI Factories and proposed AI Gigafactories. The Commission describes Gigafactories as large-scale facilities designed to support the development and training of next-generation AI models with very large computing capacity.

In July 2026, the EU launched a call to establish up to seven AI Gigafactories, backed by up to €10 billion in EU and national funding and expected to mobilise at least €20 billion in private investment.

This demonstrates an important shift.

The European governance model is no longer only about regulating AI.

It is increasingly about building the capacity required to participate in AI development.

Regulation and sovereignty are beginning to converge.

From AI Regulation to AI Industrial Policy

This convergence is visible globally.

The earlier AI policy question was:

How should governments regulate AI?

The emerging question is broader:

How should governments build the capabilities necessary to remain strategically relevant in an AI economy?

That requires infrastructure.

It requires skilled workers.

It requires research institutions.

It requires access to compute.

It requires reliable energy.

It requires data.

It requires capital.

And it requires institutional capacity.

This is why AI industrial policy has become an important part of AI governance.

Governments are not simply regulating markets.

They are attempting to shape the technological ecosystems on which future AI capabilities will depend.

Compute Is Becoming Geopolitical Power

One of the clearest examples is compute.

Advanced AI development requires enormous amounts of computational infrastructure.

Access to advanced processors therefore becomes strategically important.

This creates a direct relationship between semiconductor policy and AI policy.

If a country controls access to certain advanced chips, it can influence who can develop certain categories of AI systems.

If a country builds large-scale AI infrastructure, it can increase domestic research and industrial capacity.

If a country lacks sufficient compute, its researchers may become dependent on foreign cloud providers or external infrastructure.

This creates a new form of technological dependency.

Compute is therefore not simply an engineering resource.

It is increasingly a strategic resource.

The Infrastructure Layer

The same logic applies to data centres.

Large-scale AI requires electricity, cooling, networking, land, fibre connectivity and reliable infrastructure.

This means AI policy is increasingly becoming connected to energy policy and infrastructure planning.

The European Commission’s AI Continent strategy explicitly links AI infrastructure with computing capacity, energy and industrial competitiveness.

AI governance therefore increasingly extends beyond ministries traditionally responsible for technology.

Energy ministries matter.

Industry ministries matter.

Trade ministries matter.

Education ministries matter.

Finance ministries matter.

National-security institutions matter.

This is one reason AI governance is becoming institutionally complex.

Standards Are Also Geopolitical

Geopolitical influence is not limited to hardware.

Technical standards can shape markets and governance.

Standards determine how systems are tested.

They influence terminology.

They define documentation practices.

They can establish evaluation methodologies.

They can become embedded in procurement and regulation.

NIST, for example, continues to play a major role in US and international AI standards work. In July 2026, NIST released an initial public draft of guidance and templates for public-facing AI documentation as part of its AI standards work.

The European Union is developing standards around implementation of its AI regulatory framework.

China is developing national AI standards.

International standards organisations are attempting to create common technical approaches.

This produces an underappreciated geopolitical competition:

Who gets to define what “safe,” “trustworthy,” “secure” or “compliant” AI actually means?

Standards can become governance infrastructure.

The Country That Defines the Test May Influence the System

This is particularly important for AI evaluation.

Suppose one jurisdiction develops the dominant methodology for measuring AI safety.

Companies may begin optimising their systems around that methodology.

Regulators may adopt it.

Procurement authorities may require it.

Other countries may use it because they lack the capacity to develop their own.

Over time, an evaluation standard can become an instrument of institutional influence.

This does not mean standards are inherently geopolitical tools.

But standards inevitably shape how technological systems are understood and assessed.

That gives standards diplomacy increasing importance.

AI Safety Diplomacy

The emergence of AI safety diplomacy adds another layer.

The Bletchley AI Safety Summit, the Seoul AI Summit and the Hiroshima AI Process demonstrated that governments are increasingly discussing advanced AI through international diplomatic mechanisms.

But diplomacy operates within a competitive environment.

Countries may want cooperation on catastrophic risks while simultaneously competing over chips, talent, models and military applications.

This creates a difficult paradox.

The same countries may need each other to reduce global AI risks while viewing each other as strategic competitors.

AI governance therefore faces a cooperation problem.

Global risks do not respect national borders.

But national-security institutions naturally prioritise national interests.

The US-China Problem

This tension is especially visible in the US-China relationship.

The two countries are competing over AI capability while also confronting risks that could potentially require international coordination.

Recent reporting shows that AI is increasingly entering the strategic agenda of US-China relations, including disputes over advanced chips, model capabilities, technology controls and AI security.

The challenge is not simply competition.

It is the possibility that competition can reduce the space for cooperation.

If governments fear that sharing technical information will strengthen a strategic rival, they may become reluctant to cooperate even on common safety problems.

This creates a governance dilemma.

AI risks may be global.

Trust remains national.

Strategic Competition and Global Rules

This tension may produce several competing governance trajectories.

One possibility is deeper international cooperation.

Countries could develop common safety standards, incident-reporting systems, evaluation methodologies and crisis communication mechanisms.

Another possibility is fragmentation.

Different geopolitical blocs could develop partially separate AI ecosystems.

Standards could diverge.

Cloud infrastructure could become more geographically restricted.

Compute supply chains could become more politically controlled.

Models could be distributed through different regulatory environments.

A third possibility is selective interoperability.

Countries could remain competitors while agreeing on narrow technical standards where cooperation provides mutual benefit.

This third pathway may be particularly important.

Global AI governance does not necessarily require geopolitical consensus on everything.

It may begin with practical areas where cooperation is mutually useful.

Global South: The Strategic Question

For developing countries, geopolitical AI competition creates a different problem.

Many developing countries are not competing to build the world’s largest foundation model.

Their concern is access.

Access to compute.

Access to affordable models.

Access to infrastructure.

Access to skills.

Access to AI markets.

Access to standards.

And access to international decision-making.

This creates a risk of becoming consumers of systems designed elsewhere.

The issue is not necessarily that foreign AI systems are inherently harmful.

The deeper issue is structural dependency.

If a country depends entirely on external companies for models, cloud infrastructure, evaluation tools and technical standards, its ability to shape its own AI future may be limited.

This is why AI sovereignty should not be understood simply as building a national frontier model.

For many countries, sovereignty may mean having sufficient capacity to understand, evaluate, procure, adapt and govern AI systems.

That is a much more achievable and strategically relevant objective.

Bangladesh and the Geopolitics of AI

Bangladesh illustrates this challenge particularly well.

It is unlikely that the country’s immediate strategic priority should be to reproduce the full frontier-AI ecosystem of the United States or China.

The more fundamental question is whether Bangladesh can develop the institutional capacity required to participate intelligently in the AI economy.

That includes technical talent.

University research capacity.

AI evaluation.

Cybersecurity.

Data governance.

Public-sector procurement expertise.

Standards knowledge.

AI policy research.

Local-language AI capability.

Compute access.

And international policy participation.

Without these capabilities, Bangladesh could become primarily a policy taker.

With them, it could become a more informed participant in regional and international AI governance discussions.

This distinction matters.

AI sovereignty can begin with governance capacity.

The Geopolitics of Talent

Talent is another strategic layer.

Advanced AI depends on researchers, engineers, cybersecurity specialists, policy experts, standards professionals and technical evaluators.

Countries therefore compete not only for companies and investment but also for people.

Universities become geopolitical infrastructure.

Research institutions become strategic assets.

Migration policy becomes connected to AI strategy.

This has implications for Bangladesh.

The question is not only whether Bangladesh produces more AI programmers.

It also needs people who understand AI technology and policy simultaneously.

The emerging field requires hybrid expertise.

AI + law.

AI + cybersecurity.

AI + economics.

AI + international relations.

AI + standards.

AI + development policy.

AI + public administration.

This is precisely where Global South institutions can build distinctive expertise.

Sovereign AI

The concept of sovereign AI has emerged partly from these concerns.

Sovereign AI generally refers to the ability of a country or region to develop and operate AI capabilities in ways that reflect its own strategic interests, infrastructure, data requirements, laws and cultural context.

But sovereignty exists on a spectrum.

A country does not need to manufacture every chip, train every foundation model and operate every cloud provider to have meaningful AI sovereignty.

Instead, sovereignty can involve different combinations of:

domestic infrastructure,

trusted foreign infrastructure,

local data governance,

open models,

national research capacity,

public-sector technical expertise,

strategic procurement,

and regulatory autonomy.

This is important because complete technological independence is extremely difficult.

Strategic resilience may be more realistic than complete self-sufficiency.

The New Geopolitical Map

The AI geopolitical map is therefore more complicated than a simple US-China competition.

The United States has major strengths in frontier models, capital, cloud infrastructure and technology companies.

China combines major AI development with industrial policy, infrastructure and state-led strategic coordination.

Europe is increasingly combining regulation with technological sovereignty and infrastructure investment.

India is building AI capability around development, domestic capacity and strategic autonomy.

Japan, South Korea and Singapore are developing distinct combinations of industrial policy, safety, standards and governance.

Other countries are attempting to develop regional AI capabilities.

The result is not a single global AI model.

It is an increasingly interconnected but potentially fragmented ecosystem.

Governance Fragmentation

Fragmentation does not necessarily mean complete separation.

Different jurisdictions can still share standards.

Companies can operate across markets.

Researchers can collaborate.

International organisations can facilitate dialogue.

But regulatory divergence can increase transaction costs.

A company may need to satisfy different documentation requirements.

An AI system may need different testing approaches.

Data may be subject to different rules.

Safety requirements may vary.

Government procurement standards may differ.

This creates a growing demand for interoperability.

The Future May Be Competitive Interoperability

A useful way to think about the future is not “one global AI governance regime” versus “complete fragmentation.”

There may be a middle ground.

Countries can remain strategically competitive while building interoperable governance mechanisms.

They might agree on common terminology.

They might recognise certain evaluation methods.

They might share incident information.

They might establish crisis communication channels.

They might cooperate on AI safety research.

At the same time, they could maintain different approaches to industrial policy, national security and domestic regulation.

This could become the practical architecture of global AI governance.

Why This Matters for International Institutions

International organisations may become increasingly important because AI governance cannot be solved entirely through national policy.

The United Nations is developing mechanisms for global AI dialogue.

The OECD continues to provide policy coordination and measurement.

UNESCO addresses ethics, human rights and development.

The Council of Europe connects AI governance with human rights and the rule of law.

Standards organisations work on technical interoperability.

Each institution addresses a different layer.

The challenge is coordination.

Without coordination, the international governance ecosystem could become crowded but fragmented.

With coordination, these institutions could provide complementary functions.

The Governance Power Question

The deeper geopolitical question is therefore not simply:

Who has the best AI?

It is:

Who can shape the environment in which AI develops?

That includes the ability to influence:

compute,

chips,

cloud infrastructure,

research,

talent,

capital,

standards,

safety evaluations,

regulation,

procurement,

international institutions,

and public narratives about responsible AI.

This is why AI governance has become a geopolitical field.

Rules are not separate from power.

Rules can shape who gains access to technology, under what conditions and with what responsibilities.

The Strategic Shift

The evolution of AI governance can now be described in several stages.

The first stage focused on ethics.

The second developed international principles.

The third introduced regulation.

The fourth focused on implementation and assurance.

The fifth is increasingly concerned with frontier AI and autonomous systems.

The sixth is geopolitical.

Who controls the infrastructure?

Who defines the standards?

Who has access to compute?

Who develops the models?

Who evaluates them?

Who writes the rules?

Who participates in international governance?

These questions increasingly overlap.

Implications for Atlas AI Institute

This geopolitical transition creates a particularly strong research opportunity for Atlas.

A useful research agenda should not simply track US-China AI news.

It should examine the deeper structure of AI power.

For example:

How does compute concentration affect AI governance?

How are semiconductor controls reshaping global AI development?

How are AI standards becoming instruments of international influence?

How can developing countries avoid technological dependency without pursuing unrealistic technological autarky?

What does sovereign AI mean for smaller economies?

How should Bangladesh position itself within competing AI ecosystems?

What institutional capabilities allow a country to remain strategically autonomous while using foreign AI infrastructure?

These questions connect AI policy with geopolitics, economics, development and national capacity.

They also create an area where Global South research remains comparatively underdeveloped.

A Bangladesh-focused study could examine the country’s position across the AI value chain rather than asking only whether Bangladesh has an AI policy.

The relevant question would be:

Where does Bangladesh sit in the global AI stack?

That could become a substantial Atlas research project.

The Future of AI Governance Will Be Geopolitical

AI governance began as a question of ethics.

It became a question of principles.

Then regulation.

Then implementation.

Now it is increasingly a question of power.

This does not mean international cooperation is disappearing.

In fact, the opposite may be true.

The more strategically important AI becomes, the more countries may need mechanisms for cooperation.

But cooperation will take place within a competitive environment.

That is likely to define the next phase of AI governance.

The central challenge for the international system will be finding ways to cooperate on shared AI risks while managing competition over technology, infrastructure, security and economic power.

For developing countries, the challenge is even more fundamental:

How can they participate in shaping the rules of an AI economy they did not create?

The answer will depend less on producing one impressive AI model and more on developing durable national capacity.

Research capacity.

Technical capacity.

Regulatory capacity.

Evaluation capacity.

Diplomatic capacity.

And institutional capacity.

That is where AI governance becomes geopolitical strategy.