AI Compliance Border Walls and Platform Fragmentation

Introduction

AI compliance border walls are legal, technical, contractual, or regulatory barriers that cause an AI service to operate differently across jurisdictions, platforms, operating systems, cloud environments, app stores, or data ecosystems. Platform fragmentation occurs when those barriers divide an otherwise potentially integrated AI market into separate technical or commercial territories.

From a competition-law perspective, the central issue is not whether regulatory compliance is legitimate—it often is—but whether compliance mechanisms are designed or applied in ways that unnecessarily restrict interoperability, reinforce a dominant platform's position, raise rivals' costs, or partition markets.

The issue is becoming particularly concrete in mobile AI. In 2026, the European Commission required Google to provide competing AI services with effective interoperability with Android capabilities, including invocation mechanisms and the ability to interact with applications. The Commission's July 2026 measures also addressed access by third-party search services, including AI chatbots, to Google's search data.

1. Meaning of AI Compliance Border Walls

An AI compliance border wall can arise when a provider must create separate technical or commercial environments because of:

• data-localisation requirements;
• privacy and data-transfer restrictions;
• cybersecurity certification;
• AI-model registration or approval;
• content-moderation requirements;
• sector-specific licensing;
• export controls;
• cloud-computing restrictions;
• algorithmic transparency requirements;
• local testing or auditing;
• app-store rules;
• platform-specific API permissions;
• identity/KYC requirements; or
• restrictions on cross-border model deployment.

A legitimate regulatory requirement can nevertheless have competition consequences if compliance costs are disproportionately borne by smaller entrants or if an incumbent platform controls the technical gateway through which compliance must occur.

2. How Compliance Can Produce Platform Fragmentation

A simplified structure is:

Regulation → Compliance requirement → Technical adaptation → Fragmented architecture → Higher switching/interoperability costs → Reduced competitive pressure

For example:

AI Assistant A operates on Platform X in Country 1.
Platform X exposes APIs and system permissions there.
Country 2 imposes additional certification requirements.
Platform X creates a separate API environment.
AI Assistant A cannot transfer the same functionality across the two environments.

If the platform itself controls access to the relevant interfaces, regulatory compliance can potentially become intertwined with platform gatekeeping.

3. Competition-Law Concerns

A. Raising Rivals' Costs

Large incumbents may possess:

• established compliance departments;
• local legal teams;
• regulatory relationships;
• extensive cybersecurity infrastructure;
• existing certifications;
• large datasets;
• automated compliance systems.

A smaller AI provider may face substantially higher proportional costs.

Competition law may therefore examine whether regulatory or technical requirements create asymmetric barriers to entry.

B. Interoperability Restrictions

AI agents increasingly need access to:

• operating-system functions;
• calendars;
• messaging applications;
• payment systems;
• search;
• maps;
• cameras;
• microphones;
• device controls;
• enterprise software;
• cloud APIs.

If a dominant operating-system provider gives its own AI service privileged access while restricting rival AI systems, interoperability becomes a competition issue.

The European Commission's 2026 Android proceedings illustrate precisely this problem. Article 6(7) DMA requires effective interoperability with relevant hardware and software features, and the Commission's proceedings specifically concerned capabilities used by Google's AI services.

4. Data Fragmentation

AI competition is heavily dependent on data.

Regulatory restrictions can divide:

EU data → EU model
US data → US model
Chinese data → China-specific model
Enterprise data → private model environment

Fragmentation can reduce:

• training-data availability;
• model-learning opportunities;
• inference quality;
• cross-border network effects;
• portability;
• economies of scale.

However, data protection and cybersecurity rules can pursue legitimate objectives. Competition analysis therefore has to distinguish necessary compliance restrictions from strategically imposed restrictions that unnecessarily foreclose competitors.

5. Geographic Market Partitioning

AI platforms can technically operate worldwide while commercially functioning as multiple separate markets.

For example:

The cumulative effect may be digital market partitioning.

6. Compliance as a Competitive Moat

A dominant firm may possess sufficient resources to comply with numerous regulatory regimes simultaneously.

A startup may not.

This creates a possible compliance moat:

Higher compliance cost → fewer entrants → fewer alternatives → greater incumbent market power.

The existence of such an effect does not itself establish an antitrust violation. Authorities would ordinarily need to examine market power, causation, foreclosure, objective justification, proportionality, and the applicable statutory framework.

7. Essential-Facility and Refusal-to-Deal Issues

Where an AI platform depends upon access to an indispensable infrastructure controlled by a dominant undertaking, refusal of access can raise issues analogous to the essential-facilities doctrine.

The leading EU framework comes from:

Bronner v Mediaprint

The Court of Justice established stringent conditions for treating refusal of access as abusive, including indispensability, elimination of effective competition, and absence of objective justification.

The doctrine remains important when considering whether an AI platform's:

• API;
• operating-system functionality;
• app distribution infrastructure;
• cloud infrastructure; or
• proprietary data

is sufficiently indispensable to trigger exceptional access obligations.

Later EU case law has reaffirmed that Bronner's conditions concern infrastructure or services where access is indispensable and substitutes are unavailable.

8. Six Important Case Laws

1. Bronner v Mediaprint

Case: Oscar Bronner GmbH & Co. KG v Mediaprint Zeitungs und Zeitschriftenverlag GmbH & Co KG, C-7/97 (1998)

Principle

A dominant company does not automatically have to supply competitors. A refusal becomes potentially abusive only under stringent conditions concerning indispensability, elimination of competition, and objective justification.

Relevance to AI

An AI company seeking access to:

• proprietary APIs;
• cloud infrastructure;
• operating-system functionality; or
• critical platform interfaces

cannot automatically claim an antitrust right of access.

The case therefore provides the starting point for analysing AI interoperability border walls.

2. Microsoft Corp. v Commission

Case: Microsoft Corp. v Commission, T-201/04 (2007)

Principle

The EU courts upheld intervention concerning Microsoft's refusal to provide interoperability information to competing work-group server products.

AI relevance

This is highly relevant to AI platforms because interoperability can become a competitive parameter.

Modern AI ecosystems involve:

AI model → operating system → applications → cloud → enterprise software

If interoperability information is strategically withheld, competitors may be unable to reproduce comparable functionality.

The Microsoft case demonstrates that technical interoperability can itself have competition-law significance.

3. Google Android

Case: Google and Alphabet v Commission, T-604/18 (2022)

Principle

The General Court examined Google's Android practices involving mobile-device manufacturers, app distribution, and competitive restrictions.

AI relevance

The case demonstrates how control over a mobile operating system can influence adjacent markets.

The modern AI analogue is:

Android → system functionality → default AI assistant → user access → downstream AI competition

The significance has become even clearer in 2026, when the Commission required effective interoperability for rival AI services on Android.

4. Google Shopping

Case: Google and Alphabet v Commission (Google Shopping), T-612/17 (2021)

Principle

The EU General Court upheld the Commission's finding concerning Google's preferential treatment of its own comparison-shopping service in general search results.

AI relevance

The underlying competitive concern is relevant to AI platform design:

platform infrastructure → preferential treatment of own downstream service → reduced visibility of rivals

An AI platform might similarly privilege:

• its own chatbot;
• its own search service;
• its own agent;
• its own advertising product; or
• its own AI applications.

The legal analysis would depend on the particular conduct and applicable law.

5. Qualcomm

Case: Qualcomm Inc. v European Commission, T-235/18 (2022)

Principle

The case concerned exclusionary payments and the competitive effects of Qualcomm's conduct in the baseband-chip market.

AI relevance

It illustrates the importance of analysing exclusionary mechanisms in technologically concentrated ecosystems.

In AI markets, comparable questions could arise where an incumbent uses:

• exclusive arrangements;
• rebates;
• technical dependencies;
• preferential access; or
• contractual restrictions

to make it harder for rival AI infrastructure providers to reach customers.

6. Slovak Telekom

Cases: Slovak Telekom a.s. v Commission, C-165/19 P and C-166/19 P (2021)

Principle

The Court considered exclusionary conduct involving access to telecommunications infrastructure and the relationship between sector regulation and Article 102 TFEU.

AI relevance

This is particularly useful for understanding regulated infrastructure plus competition law.

AI ecosystems increasingly resemble infrastructure markets involving:

• cloud;
• telecommunications;
• compute;
• data centres;
• APIs;
• operating systems.

Regulation does not automatically remove competition-law scrutiny, while competition law must take account of the applicable regulatory framework.

9. Additional Relevant Authority: Lietuvos geležinkeliai

Case: European Commission v Lietuvos geležinkeliai, C-42/21 P (2023)

The case concerned the removal of railway infrastructure and the resulting restriction of access by competitors.

Its significance for AI is conceptual: where a dominant undertaking controls infrastructure needed for downstream competition, physical or technical alteration of that infrastructure can itself have exclusionary effects.

For AI, the comparable infrastructure may be virtual:

• API architecture;
• model interfaces;
• cloud environments;
• data access;
• identity systems; or
• operating-system permissions.

10. AI-Specific 2026 Development: Google Android

This is one of the clearest current examples.

In January 2026, the European Commission opened DMA specification proceedings concerning Google's obligation to provide third-party developers with effective interoperability with Android features used by Google's AI services.

The Commission subsequently proposed measures addressing:

• AI wake words;
• system-wide invocation;
• contextual information;
• application interaction;
• task execution;
• relevant hardware/software resources.

In July 2026, the Commission adopted binding specification measures. The stated objective was to enable competing AI services to access Android capabilities on an equal footing with Google's AI services.

This is particularly important because it demonstrates a shift from traditional ex post antitrust enforcement toward ex ante interoperability regulation.

11. Regulatory Fragmentation vs Anticompetitive Fragmentation

These concepts should not be confused.

Regulatory fragmentation

Different jurisdictions legitimately impose different:

• privacy rules;
• safety standards;
• cybersecurity requirements;
• AI governance systems;
• data-transfer rules.

Anticompetitive fragmentation

A dominant platform potentially uses:

• incompatible APIs;
• discriminatory access;
• technical restrictions;
• contractual barriers;
• discriminatory certification;
• self-preferencing;

to make rivals less competitive.

The first can be legitimate public regulation. The second can potentially attract competition-law scrutiny.

12. Objective Justification and Security

AI platforms can legitimately argue that interoperability creates:

• cybersecurity risks;
• privacy risks;
• model-extraction risks;
• prompt-injection vulnerabilities;
• malicious automation;
• unauthorized access;
• data leakage.

Google, for example, publicly argued in 2026 that deep system-level AI-agent interoperability could create privacy and security risks.

Consequently, competition authorities must distinguish between:

genuine security restrictions

and

security claims used as a pretext for exclusion.

This makes technical evidence particularly important.

13. Platform Fragmentation and Switching Costs

Fragmentation can also increase switching costs.

A user might have:

AI Agent A + Android + Google data + Google apps

and switching to another ecosystem could require:

• exporting data;
• recreating preferences;
• changing authentication;
• changing applications;
• losing contextual information;
• retraining workflows.

The greater the ecosystem dependency, the greater the potential competitive importance of data portability and interoperability.

The DMA expressly addresses data portability as well as interoperability, reflecting the importance of these mechanisms to contestability in mobile ecosystems.

14. Competition Concerns in AI Compliance Architecture

Authorities may therefore examine:

1. Market definition

Is the relevant market:

• AI assistants?
• foundation models?
• AI operating systems?
• cloud AI?
• AI APIs?
• enterprise AI agents?

2. Market power

Does the platform control an indispensable gateway?

3. Foreclosure

Are rivals prevented from reaching users or data?

4. Interoperability

Are equivalent technical capabilities offered to rivals?

5. Discrimination

Does the platform treat its own AI service more favourably?

6. Data access

Can competitors obtain necessary data on reasonable terms?

7. Regulatory justification

Is the restriction genuinely required by law, security, privacy, or safety?

8. Proportionality

Could the legitimate objective be achieved through a less restrictive mechanism?

15. Compliance Border Walls and M&A

Fragmentation also matters in AI mergers.

Suppose an AI company acquires:

model + cloud + data + distribution + compliance infrastructure

The transaction could potentially increase vertical integration.

Competition authorities may examine whether the merged firm could:

• deny API access;
• degrade interoperability;
• restrict data portability;
• impose discriminatory compliance requirements;
• bundle AI with cloud services;
• foreclose independent AI developers.

Thus, compliance infrastructure itself can become strategically important merger infrastructure.

16. Remedies

Possible competition-law remedies include:

Interoperability

Require equivalent access to relevant APIs and operating-system functionality.

Data portability

Enable users and businesses to transfer relevant data between AI services.

Non-discrimination

Prevent preferential treatment of the platform's own AI.

FRAND-style access

Where legally appropriate, require fair, reasonable and non-discriminatory access conditions.

Transparency

Require documentation of:

• API restrictions;
• technical standards;
• access criteria;
• security requirements.

Compliance separation

Require regulatory compliance mechanisms to be administered in a way that does not unnecessarily discriminate against competing services.

17. Key Legal Tension

The central tension can be represented as:

AI SAFETY / PRIVACY / SECURITY

⬇

Compliance requirements

⬇

Interoperability restrictions

⬇

Platform fragmentation

⬇

Potential entry barriers

versus

Competition / innovation / interoperability / consumer choice

The legal task is not simply to eliminate compliance barriers. It is to determine which restrictions are necessary and proportionate and which may unnecessarily protect or reinforce market power.

Conclusion

AI compliance border walls and platform fragmentation represent an emerging intersection of competition law, digital regulation, data governance, cybersecurity and AI regulation.

The established cases—Bronner, Microsoft, Google Android, Google Shopping, Qualcomm, Slovak Telekom and Lietuvos geležinkeliai—provide different doctrinal tools for analysing access, interoperability, exclusion, infrastructure dependence and regulatory interaction.

The most significant contemporary development is the EU's 2026 Android AI interoperability intervention. The Commission's measures demonstrate that AI competition is increasingly concerned not merely with the quality of models, but with who controls the technical gateways through which AI services reach users, applications, data and devices.