Global AI Competition Law Harmonization Trajectories

Introduction

Global AI competition-law harmonization refers to the gradual convergence of competition/antitrust rules, enforcement approaches, institutional practices, and regulatory principles governing artificial-intelligence markets across jurisdictions. It does not necessarily mean that countries will adopt one identical AI antitrust statute. Rather, harmonization is likely to develop through common principles concerning AI compute, data, foundation models, cloud infrastructure, algorithms, AI applications, distribution channels, interoperability, acquisitions, vertical integration, and algorithmic coordination.

AI markets create unusually strong reasons for international convergence because the relevant markets are inherently cross-border. A foundation model may be trained using data from several countries, run on cloud infrastructure located elsewhere, accessed through APIs globally, and incorporated into applications distributed worldwide.

At the same time, complete harmonization is difficult because the EU, United States, United Kingdom, Germany, China, Japan, India and other jurisdictions have different competition-policy traditions, institutional structures and regulatory objectives.

The principal trajectory is therefore likely to be functional harmonization rather than uniformity.

1. Meaning of AI Competition-Law Harmonization

AI competition-law harmonization involves convergence around several questions:

1. How should AI-relevant markets be defined?
2. When does control over data constitute market power?
3. Can compute infrastructure constitute an essential input?
4. When is an AI-related acquisition anticompetitive even if traditional merger thresholds are not met?
5. How should algorithmic coordination be treated?
6. How should self-preferencing by AI platforms be analysed?
7. What interoperability and data-access obligations can competition authorities impose?
8. How should competition law interact with AI regulation, privacy, cybersecurity and intellectual-property law?
9. How should regulators cooperate where AI conduct affects multiple jurisdictions?

The central movement is from a traditional conception of competition based principally on price and output toward an analysis incorporating:

• innovation;
• access to compute;
• data;
• model quality;
• interoperability;
• switching costs;
• ecosystem effects;
• attention;
• algorithmic dependency;
• distribution;
• technical standards; and
• control over AI infrastructure.

2. Why Global Harmonization Is Becoming Necessary

AI markets have several characteristics that make purely national enforcement inadequate.

A. Cross-border foundation models

A single foundation model can serve users in dozens of jurisdictions. An exclusionary agreement concerning model distribution can therefore have international effects.

B. Concentration of compute

Advanced AI development depends heavily on sophisticated GPUs, specialized accelerators, cloud infrastructure and large-scale data centres.

Control over these inputs may create bottlenecks analogous to traditional essential facilities.

C. Global cloud markets

AI developers frequently depend upon cloud providers for training and inference.

A cloud provider that simultaneously operates an AI model or application may possess incentives to:

• favour its own models;
• restrict competitors' access;
• impose discriminatory technical conditions;
• bundle compute with AI services; or
• make switching difficult.

D. Global AI acquisitions

AI companies may be acquired before they become large enough to trigger conventional merger thresholds.

This creates a killer-acquisition / nascent-competition problem.

E. Algorithmic coordination

AI agents can independently process enormous quantities of market information and potentially produce parallel pricing or allocation decisions.

This raises the possibility of coordination without a conventional human agreement.

3. First Trajectory: Convergence Around Digital-Market Principles

The first harmonization trajectory is the increasing recognition that traditional competition-law concepts must be adapted to digital markets.

Authorities increasingly examine:

• network effects;
• data advantages;
• economies of scale;
• multi-sided markets;
• switching costs;
• interoperability;
• default settings;
• ecosystem control; and
• vertical integration.

These concepts are particularly relevant to AI because the competitive position of a foundation-model provider may depend not merely upon the model itself but upon its control over the entire AI stack.

The stack may be represented as:

Semiconductors → Compute → Cloud → Data → Foundation Model → API → Applications → Distribution → Users

A competition authority increasingly has to ask whether control at one level allows foreclosure at another.

4. Second Trajectory: Harmonization of AI Merger Scrutiny

A major area of convergence is likely to be merger control involving AI firms.

Traditional merger thresholds often depend upon turnover or transaction value. But AI startups may possess:

• valuable technology;
• engineers;
• datasets;
• intellectual property;
• users;
• algorithms;
• research capabilities;

without generating substantial revenue.

Consequently, authorities are increasingly interested in:

Killer acquisitions

A dominant technology company acquires an emerging AI competitor before it becomes a serious competitive threat.

Acqui-hiring

The acquiring company nominally acquires personnel rather than the business, but effectively eliminates independent competitive development.

Strategic AI partnerships

A transaction may not be a conventional acquisition but may nevertheless give one firm substantial control over:

• model development;
• compute;
• distribution;
• APIs;
• technical personnel; or
• intellectual property.

Minority investments

Even a non-controlling investment can potentially produce:

• information exchange;
• board influence;
• commercial dependency;
• reduced incentives to compete; or
• preferential access.

5. Third Trajectory: Convergence Around Data as a Competitive Asset

AI competition law increasingly treats data as an important source of competitive advantage.

However, the legal question is not simply whether a company possesses "a lot of data."

Authorities must investigate:

• uniqueness;
• scale;
• quality;
• freshness;
• exclusivity;
• substitutability;
• access conditions;
• portability;
• interoperability; and
• whether competitors can realistically reproduce the dataset.

This produces a convergence between:

Competition law + data protection + data governance + AI regulation.

A dominant AI platform could potentially use its data advantage to:

1. improve its model;
2. attract more users;
3. collect more data;
4. further improve the model; and
5. increase barriers to entry.

This can create a data–model–user feedback loop.

6. Fourth Trajectory: Compute as a Potential Bottleneck

One of the most important future harmonization questions concerns AI compute.

Advanced models require enormous computational resources.

Where access to suitable compute is limited, competition authorities may examine whether dominant infrastructure providers can:

• discriminate between AI developers;
• reserve capacity for affiliated models;
• impose excessive contractual restrictions;
• bundle compute with unrelated services;
• restrict portability;
• impose high switching costs; or
• engage in discriminatory allocation during shortages.

The analytical framework could eventually resemble essential-input or essential-facility doctrines.

However, authorities must be cautious.

Scarcity alone does not establish an essential facility.

The analysis must establish factors such as:

• market power;
• indispensability;
• lack of reasonable alternatives;
• foreclosure;
• competitive harm; and
• proportionality of the proposed remedy.

7. Fifth Trajectory: Interoperability as a Global Competition Principle

Interoperability is becoming particularly important in AI.

Potential forms include:

• model interoperability;
• API interoperability;
• data portability;
• agent interoperability;
• identity portability;
• cloud portability;
• model switching;
• application compatibility.

A dominant AI ecosystem could potentially make switching difficult by designing technical architecture that causes users to become dependent upon its:

• APIs;
• data formats;
• model-specific tools;
• plugins;
• agent systems;
• cloud infrastructure.

Competition law may therefore increasingly treat technical interoperability as a competitive parameter.

8. Sixth Trajectory: Algorithmic Collusion

AI creates a particularly difficult form of coordination.

Traditional cartel law usually focuses upon:

• agreements;
• concerted practices;
• communication;
• intentional coordination.

AI systems can complicate this framework because autonomous systems may:

• observe competitors' prices;
• predict reactions;
• modify prices;
• experiment with strategies;
• optimize for long-term profit; and
• respond dynamically to competitors.

The critical legal distinction will remain:

Does autonomous parallel behaviour amount to legally attributable coordination?

Competition-law harmonization will probably move toward clearer rules concerning:

• human responsibility;
• algorithmic design;
• foreseeable coordination;
• information exchange;
• common algorithms;
• pricing software;
• monitoring systems; and
• liability of developers and deployers.

9. Seventh Trajectory: Self-Preferencing by AI Platforms

An integrated AI platform may operate simultaneously as:

• infrastructure provider;
• foundation-model provider;
• API provider;
• application provider;
• marketplace;
• search/discovery system; and
• distribution channel.

This produces incentives for self-preferencing.

For example, a platform could potentially rank its own AI application above competing applications.

The competition-law question becomes whether the platform is merely exercising legitimate product design or using dominance in one layer to distort competition in another.

This is particularly relevant to the EU's Article 102 TFEU framework and broader digital-market regulation.

10. Eighth Trajectory: Increasing Merger-Remedy Convergence

Future international cooperation may increasingly produce similar remedies.

Possible remedies include:

Structural remedies

• divestiture;
• separation of assets;
• restrictions on ownership.

Behavioural remedies

• non-discrimination;
• access obligations;
• interoperability;
• licensing;
• data access;
• transparency;
• contractual restrictions.

Governance remedies

• independent compliance monitoring;
• audit mechanisms;
• reporting obligations;
• firewalls;
• information-sharing restrictions.

Technical remedies

• API access;
• portability;
• interoperability;
• model switching;
• standardized technical interfaces.

The likely trajectory is toward technology-neutral remedies combined with AI-specific technical obligations.

11. Ninth Trajectory: EU–US Convergence Without Complete Uniformity

The EU and US historically have different competition-law philosophies.

The EU traditionally gives greater importance to:

• market structure;
• fairness;
• economic dependence;
• exclusionary conduct;
• regulatory objectives.

US antitrust has traditionally placed greater emphasis upon:

• consumer welfare;
• economic efficiency;
• effects on competition;
• demonstrable competitive harm.

AI is nevertheless producing substantial convergence because both systems increasingly confront:

• digital gatekeepers;
• platform concentration;
• AI acquisitions;
• cloud dependency;
• data advantages;
• interoperability problems.

The convergence is therefore likely to occur through shared economic analysis, even where statutory standards remain different.

12. Tenth Trajectory: EU–UK Regulatory Convergence and Divergence

The UK increasingly occupies a distinctive position.

Its competition regime combines:

• Competition Act 1998;
• Enterprise Act 2002;
• Digital Markets, Competition and Consumers Act 2024;
• CMA enforcement;
• sectoral regulation.

The UK is likely to converge with the EU on issues such as:

• digital gatekeeper power;
• interoperability;
• data access;
• AI mergers;
• platform self-preferencing.

But it may simultaneously develop independent approaches suited to the UK digital economy.

Thus, regulatory interoperability rather than literal legal identity is likely to be the dominant model.

13. Eleventh Trajectory: China and State-Centred AI Competition Regulation

China presents a different institutional model.

Chinese competition regulation increasingly interacts with:

• industrial policy;
• cybersecurity;
• data governance;
• platform regulation;
• algorithm regulation;
• strategic technology policy.

Consequently, international harmonization involving China is unlikely to produce complete substantive uniformity.

Nevertheless, there can be convergence around concepts such as:

• platform dominance;
• algorithmic discrimination;
• abusive exclusivity;
• interoperability;
• unfair treatment of smaller firms;
• concentration in strategic technology markets.

The major difference concerns the relationship between competition objectives and state economic/industrial objectives.

14. Twelfth Trajectory: Developing Economies and AI Market Access

Global harmonization also has a significant developing-country dimension.

AI infrastructure may be concentrated in a small number of companies and countries.

This creates concerns regarding:

• access to compute;
• cloud dependency;
• data concentration;
• technology licensing;
• AI model availability;
• technical standards;
• cross-border data flows.

Competition law may therefore increasingly become connected with digital sovereignty and market-access concerns.

India and other developing economies may seek to prevent global AI concentration from becoming a structural barrier to domestic AI development.

15. Important Case Laws

The following cases provide the principal doctrinal foundations for the emerging global AI competition-law framework.

1. Microsoft Corp. v. Commission — EU

The Microsoft litigation is fundamental to understanding leveraging, interoperability and tying.

The European Commission and EU courts examined Microsoft's ability to leverage dominance in one market into adjacent markets.

AI significance

The case provides a framework for examining whether a dominant AI/cloud ecosystem could:

• tie AI models to cloud services;
• restrict interoperability;
• deny competitors technical information;
• leverage infrastructure dominance into application markets.

It therefore remains highly relevant to AI ecosystem regulation.

2. Google Shopping — Google and Alphabet v Commission

The EU Google Shopping litigation concerned preferential treatment of Google's own comparison-shopping service in search results.

AI significance

Its broader relevance lies in the concept of self-preferencing by a vertically integrated digital platform.

The same logic may become relevant where an AI platform:

• owns the infrastructure;
• operates the foundation model;
• controls the interface; and
• gives preferential treatment to its own applications.

It is therefore a major reference point for future AI-platform cases.

3. Google Android — Google and Alphabet v Commission

The Android litigation concerned Google's contractual practices concerning mobile operating systems, search and applications.

AI significance

The case illustrates how dominance in an important technological ecosystem can potentially be extended into neighbouring markets through:

• contractual restrictions;
• defaults;
• tying;
• distribution arrangements.

This is highly relevant to AI ecosystems where a dominant model could be distributed through:

• operating systems;
• browsers;
• cloud platforms;
• app stores; or
• devices.

4. Qualcomm v Commission

The Qualcomm litigation concerned exclusionary conduct involving payments and incentives in the semiconductor sector.

AI significance

Semiconductors are an upstream bottleneck for AI.

The case provides useful analytical foundations for examining whether a dominant AI-chip supplier could use:

• exclusivity;
• conditional payments;
• rebates;
• contractual incentives;

to prevent competing AI accelerator suppliers from gaining market access.

It therefore links traditional Article 102 doctrine with emerging AI-compute competition.

5. Intel v Commission

Intel concerned rebates and exclusionary strategies by a dominant undertaking.

AI significance

The case is important for analysing conditional rebates where a dominant AI infrastructure provider might offer advantageous pricing only if customers:

• purchase exclusively;
• allocate most workloads to its infrastructure;
• use affiliated AI services.

The judgment also illustrates the importance of rigorous effects analysis.

6. United States v. Microsoft Corp.

The US Microsoft antitrust litigation remains one of the most important precedents concerning technology-platform dominance.

The case addressed Microsoft's use of its operating-system dominance to protect its position against emerging competitive threats.

AI significance

It provides an important conceptual analogy for situations where an established technology platform might use control over an existing distribution ecosystem to restrict an emerging AI technology.

Its importance is particularly strong for understanding:

• platform leverage;
• exclusion;
• distribution control;
• technological integration; and
• nascent competition.

7. FTC v. Qualcomm

The Qualcomm litigation in the United States examined licensing and competitive conduct in the cellular-chip ecosystem.

AI significance

It demonstrates the difficulty of applying antitrust law to technology licensing and vertically integrated infrastructure markets.

AI competition disputes may similarly involve:

• semiconductor IP;
• model licensing;
• cloud infrastructure;
• APIs;
• standards;
• patent portfolios.

The case is especially useful for understanding the limits of using antitrust law to regulate complex licensing arrangements.

8. Ohio v. American Express

The US Supreme Court's decision concerning American Express is important for two-sided platform analysis.

AI significance

AI platforms increasingly operate as multi-sided ecosystems connecting:

• users;
• developers;
• advertisers;
• application providers;
• data suppliers;
• cloud customers.

The case reinforces the importance of correctly defining the competitive framework before assessing platform conduct.

16. German Influence on Harmonization

German competition law is particularly important because of its strong focus on economic power and digital ecosystems.

The German approach has increasingly emphasized the significance of:

• intermediaries;
• data;
• network effects;
• strategic importance;
• access to competition-relevant resources;
• ecosystem power.

The modern German framework, particularly GWB §19a, provides an important model for dealing with firms of paramount significance across markets.

Its conceptual importance for AI is substantial because an AI company may exercise power across multiple interconnected markets without holding a conventional monopoly in every individual market.

17. The Emerging "AI Stack" Competition Model

A significant future development is likely to be analysis of competition at each level of the AI stack.

This represents a shift from market-by-market analysis toward ecosystem and stack analysis.

18. Institutional Harmonization

Substantive law is only one component of harmonization.

Competition authorities will increasingly need mechanisms for:

• information sharing;
• coordinated investigations;
• merger review;
• dawn-raid coordination;
• economic analysis;
• technical expertise;
• remedy coordination;
• cross-border monitoring.

Important institutional networks include:

• International Competition Network;
• OECD;
• European Competition Network;
• bilateral cooperation between competition authorities;
• regional competition networks.

The likely future model is coordinated enforcement rather than a single global AI competition regulator.

19. Convergence of Economic Methodology

Global regulators are also likely to converge around economic tools.

These may include:

Market definition

AI markets may require analysis of:

• model quality;
• functionality;
• latency;
• reliability;
• privacy;
• interoperability;
• price;
• compute requirements.

Counterfactual analysis

Authorities may ask:

What would the AI market have looked like without the alleged exclusionary conduct?

Entry analysis

They may examine:

• compute availability;
• capital requirements;
• access to data;
• engineering talent;
• model training costs;
• distribution.

Innovation analysis

AI competition may require evaluating whether conduct reduces:

• model innovation;
• application innovation;
• safety innovation;
• research diversity.

20. Tension Between AI Regulation and Competition Law

One of the biggest harmonization challenges is the relationship between competition law and AI regulation.

AI regulation may require:

• safety testing;
• documentation;
• conformity assessment;
• cybersecurity;
• risk management;
• transparency.

But these requirements can unintentionally raise barriers to entry.

For example, if compliance costs are very high, large AI firms may be able to absorb them more easily than startups.

Therefore:

AI safety regulation can itself affect competition.

Global regulators will increasingly need to conduct competition assessments of AI regulation.

21. Privacy–Competition–AI Convergence

AI competition regulation will also intersect with privacy law.

Data access can improve competition, but unrestricted data access may undermine:

• privacy;
• confidentiality;
• cybersecurity;
• intellectual property.

Accordingly, regulators will increasingly ask:

How can data be made competitively accessible without eliminating legitimate privacy protections?

Potential solutions include:

• anonymisation;
• secure data environments;
• federated access;
• trusted data intermediaries;
• purpose limitation;
• controlled API access.

22. Intellectual Property and AI Competition

AI markets heavily depend upon:

• patents;
• copyright;
• model weights;
• training datasets;
• trade secrets;
• software licences.

Competition law must therefore distinguish between legitimate IP protection and strategic exclusion.

Potential concerns include:

• refusal to license;
• discriminatory licensing;
• patent aggregation;
• exclusive model licences;
• restrictive training-data licences;
• technological interoperability restrictions.

This creates another major area for international convergence.

23. Toward an International AI Competition Principles Framework

A future harmonized framework could contain principles such as:

Principle 1 — Non-discrimination

Dominant AI infrastructure providers should not discriminate against competing AI services without legitimate justification.

Principle 2 — Interoperability

Dominant ecosystems should not unnecessarily prevent technically feasible switching and interoperability.

Principle 3 — Competitive access

Control over indispensable AI inputs should not be abused to exclude competitors.

Principle 4 — Merger scrutiny

AI acquisitions should be assessed for their effect on innovation and potential competition.

Principle 5 — Algorithmic accountability

Competition responsibility should not disappear merely because decision-making is automated.

Principle 6 — Procedural transparency

AI systems used in competition enforcement should be subject to appropriate oversight.

Principle 7 — Regulatory cooperation

Cross-border AI investigations should involve appropriate international coordination.

24. Major Obstacles to Harmonization

A. Different legal standards

Article 102 TFEU, US Sherman Act §2, UK Competition Act 1998 and China's Anti-Monopoly Law do not operate identically.

B. Different policy objectives

Some jurisdictions prioritize:

• consumer welfare;

others give greater weight to:

• market structure;
• fairness;
• innovation;
• economic democracy;
• national technological capacity.

C. Digital sovereignty

States may resist rules requiring extensive foreign access to domestic AI infrastructure or data.

D. National-security considerations

AI has strategic applications involving:

• defence;
• intelligence;
• cybersecurity;
• critical infrastructure.

Competition policy can therefore conflict with national-security policy.

E. Divergent privacy rules

Data-access remedies may conflict with different privacy regimes.

F. Enforcement asymmetry

Large jurisdictions may have significantly greater technical and institutional resources than smaller competition authorities.

25. Likely Future Trajectory

The most plausible development is layered harmonization.

Stage 1 — Conceptual convergence

Authorities develop similar understandings of:

• AI markets;
• compute;
• data;
• algorithms;
• platform power.

Stage 2 — Enforcement convergence

Authorities begin investigating similar conduct:

• AI mergers;
• exclusivity;
• tying;
• self-preferencing;
• compute foreclosure;
• algorithmic coordination.

Stage 3 — Procedural convergence

Authorities coordinate:

• investigations;
• evidence;
• economists;
• technical experts;
• remedies.

Stage 4 — Remedy convergence

Similar obligations emerge concerning:

• interoperability;
• access;
• non-discrimination;
• portability;
• transparency.

Stage 5 — International principles

OECD, ICN, G20 and regional institutions may gradually develop common AI competition principles.

Stage 6 — Regulatory interoperability

Rather than one global statute, jurisdictions may design their systems so that their competition authorities can recognize and coordinate with one another.

26. Possible Global AI Competition Governance Architecture

A future architecture could look like:

International Principles
↓
OECD / ICN / G20 Guidelines
↓
Regional Frameworks
EU / UK / ASEAN / other regional regimes
↓
National Competition Authorities
CMA / European Commission / FTC / DOJ / Bundeskartellamt / CCI / others
↓
Sector Regulators
AI / telecommunications / cloud / financial / energy regulators
↓
Technical Standards and Auditing Bodies
↓
AI Firms and Platforms

This would create horizontal coordination without requiring a single global competition authority.

27. Significance for India

India is particularly important to this harmonization trajectory because its AI market combines:

• a large digital-user base;
• expanding domestic AI development;
• major platform ecosystems;
• cloud dependence;
• significant data resources;
• developing AI infrastructure.

The Competition Commission of India may increasingly confront questions involving:

• AI platform dominance;
• algorithmic pricing;
• cloud–AI bundling;
• data advantages;
• digital ecosystems;
• AI acquisitions;
• self-preferencing;
• interoperability;
• discriminatory access to compute.

Indian competition policy may consequently develop as a bridge between global competition standards and domestic digital-development priorities.

28. Overall Legal Assessment

The most important point is that global AI competition-law harmonization is unlikely to mean identical legislation.

Instead, convergence is likely to occur around a common vocabulary:

market power + data + compute + algorithms + interoperability + innovation + ecosystem effects + cross-border enforcement.

The EU may continue emphasizing structural and gatekeeper concerns; the US may retain a stronger effects-oriented antitrust tradition; the UK may develop a flexible digital-markets framework; Germany may continue emphasizing cross-market economic power; China may integrate competition regulation with broader digital governance; and India may combine competition principles with digital-development considerations.

Yet these different systems can still converge around common enforcement problems.

Conclusion

Global AI competition-law harmonization is moving toward functional convergence rather than complete legal uniformity.

The central transformation is from traditional antitrust analysis of individual products toward regulation of AI ecosystems and strategic bottlenecks.

The most important future areas of convergence are likely to be:

1. AI merger and acquisition scrutiny;
2. compute and cloud access;
3. data concentration;
4. foundation-model dominance;
5. self-preferencing;
6. algorithmic coordination;
7. interoperability and portability;
8. AI licensing and IP-related exclusion;
9. cross-border enforcement cooperation;
10. coordinated remedies for systemic AI platforms.

The long-term trajectory is therefore best described as:

National Competition Law → Digital Competition Law → AI Competition Law → Cross-Border Enforcement Cooperation → International AI Competition Principles.

The six-plus leading cases—Microsoft, Google Shopping, Google Android, Qualcomm, Intel, US Microsoft, FTC v Qualcomm, and Ohio v American Express—provide important doctrinal building blocks for this evolution, even though most pre-date modern generative AI.

Ultimately, the emerging global principle is likely to be that AI innovation should remain contestable at every critical layer of the AI stack, while competition authorities cooperate sufficiently to prevent firms from converting control over data, compute, models, interfaces or distribution into durable global exclusion.