API Economy, AI Orchestration and Dominance Issues

1. Introduction

The API economy refers to the commercial environment in which software, platforms, applications, data and artificial-intelligence services interact through Application Programming Interfaces (APIs). APIs can become strategically important because they determine who can access functionality, data, models, computing resources, users and complementary services.

The emergence of AI orchestration adds another layer. An AI-orchestration platform may decide:

• which AI model or agent receives a task;
• which API is called;
• what data is supplied to the model;
• which application performs the resulting action;
• how competing AI services are ranked or selected;
• what permissions an AI agent receives;
• how much developers pay for API calls;
• whether users can switch between AI providers; and
• whether the platform's own AI receives privileged access.

Competition concerns therefore arise when a firm controlling a critical API or orchestration layer can use that control to extend market power into adjacent AI, software, cloud, search, advertising, payments or application markets.

The OECD has specifically identified control over compute, proprietary data and model interfaces such as APIs as potential mechanisms through which vertically integrated AI firms can privilege affiliated services, restrict interoperability and increase switching costs.

Importantly, most existing cases do not concern AI orchestration in the modern sense. The established cases provide analogies concerning interoperability, tying, self-preferencing, refusal to supply, data access, algorithmic coordination and ecosystem leverage.

2. Meaning of AI Orchestration Dominance

A. Traditional API economy

A simplified structure is:

Platform → API → Developers → Applications → Consumers

The API may provide:

• payment functionality;
• identity verification;
• maps;
• cloud computing;
• search;
• advertising;
• data;
• AI models;
• authentication;
• messaging;
• device functionality.

If the API provider becomes indispensable, it can potentially influence downstream competition.

B. AI orchestration economy

AI orchestration creates a more complicated structure:

User → Orchestrator → AI Model/Agent → API → Data/Application → Result

An orchestrator can potentially control the entire chain.

For example, an enterprise AI assistant might determine whether a user's request is handled by:

• its own AI model;
• a rival foundation model;
• a third-party specialist agent;
• a cloud service;
• an external application;
• a payment API; or
• a search provider.

The selection mechanism itself can therefore become a competitive bottleneck.

3. When Does API Control Become Market Power?

API ownership alone does not establish dominance.

Competition authorities would ordinarily examine factors such as:

1. Market share

What proportion of relevant transactions, developers, users or API calls pass through the platform?

2. Network effects

Does more usage attract more developers, which attracts more users, which in turn generates more usage?

3. Switching costs

Can developers realistically move from one API to another?

4. Data advantages

Does the API provider obtain proprietary data unavailable to competitors?

5. Technical dependency

Would downstream firms lose important functionality if access were withdrawn?

6. Interoperability

Can competing APIs, models and agents interact with the ecosystem?

7. Vertical integration

Does the API provider simultaneously operate:

• cloud infrastructure;
• foundation models;
• AI assistants;
• application stores;
• search engines;
• advertising services;
• productivity software?

8. Multi-homing

Can developers and consumers easily use several competing providers simultaneously?

4. Major Dominance Issues

A. Self-preferencing in AI orchestration

The orchestrator may systematically prefer its own AI model.

For example:

User request → AI orchestrator → proprietary AI model automatically selected

while competing models are:

• ranked lower;
• given less context;
• subjected to slower API access;
• excluded from certain device functions; or
• charged more.

The competition concern is particularly strong where the orchestration platform controls the decision-making layer through which customers reach competing AI services.

The EU's 2026 Android interoperability measures illustrate the importance of this issue: the Commission required effective access for competing AI services to Android features that Google's own AI services could use, including functionality enabling AI assistants to interact with applications.

5. Refusal to Provide API Access

A dominant platform may refuse access to an API that competitors need.

Potential theories include:

• refusal to supply;
• essential-facility-type theories;
• exclusionary conduct;
• raising rivals' costs;
• foreclosure of adjacent markets.

However, competition law generally does not impose a universal obligation on dominant firms to share every proprietary technology.

The exceptional nature of compulsory access is demonstrated by the Bronner and IMS Health lines of authority.

6. API Interoperability Restrictions

A platform may technically permit API access while making interoperability commercially or technically ineffective.

Examples include:

• incomplete documentation;
• delayed access;
• restrictive authentication;
• rate limits;
• discriminatory permissions;
• missing functionality;
• incompatibility with competing agents;
• discriminatory pricing;
• preferential access to the platform's own AI.

This is especially significant because modern AI assistants increasingly need deep access to operating-system and application functionality.

The EU's DMA expressly addresses this problem. Article 6(7) requires designated gatekeepers to provide effective interoperability with relevant hardware and software features controlled by their operating systems.

7. API Tying and Bundling

A dominant firm may condition access to one API or service on purchasing another service.

Examples:

Cloud API + AI model

Operating system + proprietary AI assistant

Enterprise software + AI copilot

App store + payment API

Search API + advertising services

The legal question is whether the bundle or tying arrangement forecloses competitors and reduces consumer or developer choice.

8. Data Foreclosure

AI orchestration depends heavily on data.

An orchestrator may possess:

• user interaction data;
• search data;
• application data;
• API-call data;
• transaction data;
• behavioural data;
• model-performance data.

The platform may use this information to improve its own AI while preventing competitors from obtaining comparable data.

This can create a feedback loop:

More users → more data → better AI → better orchestration → more users → more data

The European Commission's 2026 AI interoperability work specifically recognized access to search data as a competitive issue because Google Search can collect such data at scale.

9. Discriminatory API Pricing

An API provider may charge:

• its own AI business: low internal cost;
• independent AI firms: high API charges.

Potentially problematic practices include:

Margin squeeze

A vertically integrated firm charges downstream competitors enough for API access that they cannot profitably compete with the provider's downstream business.

Differential pricing

The dominant firm's own subsidiary receives preferential rates.

Minimum commitments

Competitors are required to purchase large API volumes.

Dynamic pricing

API prices increase when a rival's usage increases.

10. Switching Barriers

AI orchestration can make switching particularly difficult.

A developer may have to change:

• API calls;
• authentication;
• data formats;
• prompts;
• model specifications;
• safety systems;
• monitoring;
• databases;
• application architecture.

Consequently:

Technical compatibility costs + data migration costs + contractual restrictions = switching barrier

High switching costs can strengthen incumbent market power even where several theoretical competitors exist.

11. Algorithmic Coordination

AI orchestration may also facilitate coordination among competitors.

Suppose competing AI agents access a common orchestration infrastructure that:

• monitors prices;
• observes demand;
• recommends pricing;
• communicates market information;
• automatically adjusts offers.

The issue changes from unilateral exclusion to possible concerted practices or algorithmic coordination.

The relevant legal principles can be found in cases concerning computerized coordination and information exchange.

12. Six Important Case Laws

1. Microsoft Corp. v Commission, Case T-201/04

Facts

The European Commission found that Microsoft had abused its dominant position through conduct involving interoperability information and tying.

The General Court substantially upheld the Commission's findings concerning Microsoft's failure to provide interoperability information and its tying of Windows Media Player.

Principle

The case is important because it demonstrates that control over technical interoperability can have competition-law consequences when a dominant firm's conduct prevents competitors from competing effectively in neighbouring markets.

Application to AI orchestration

An AI platform could potentially control APIs necessary for competing AI agents to interact with:

• operating systems;
• enterprise applications;
• databases;
• cloud services.

If access is withheld or technically degraded, Microsoft provides an important analogy.

Relevance: interoperability foreclosure and tying.

13. IMS Health GmbH & Co. KG v NDC Health GmbH, Case C-418/01

Facts

IMS Health controlled a data structure used for pharmaceutical sales information. A rival sought access to the structure.

The Court of Justice established stringent conditions concerning when refusal by a dominant undertaking to license intellectual property could constitute abuse.

Principle

Compulsory access requires exceptional circumstances, including considerations concerning indispensability and elimination of effective competition.

Application to APIs

An AI API might become legally significant if:

1. access is indispensable;
2. there is no realistic substitute;
3. refusal excludes effective competition;
4. access cannot be reasonably duplicated; and
5. refusal lacks sufficient justification.

Therefore, not every API becomes an essential facility merely because competitors would like access to it.

14. Bronner v Mediaprint, Case C-7/97

Facts

Oscar Bronner sought access to Mediaprint's newspaper distribution system.

The Court imposed a demanding test before a dominant undertaking could be required to provide access to infrastructure.

Principle

A refusal to supply becomes an abuse only in exceptional circumstances. The infrastructure must generally be indispensable, and duplication must not be realistically feasible.

Application to AI orchestration

Suppose an AI orchestrator controls the only technically viable mechanism through which applications can:

• invoke certain device functions;
• access a critical dataset;
• communicate with particular enterprise software.

Bronner becomes relevant to determining whether compulsory access is justified.

Importance

The case prevents competition law from becoming a general requirement that successful technology companies must open every proprietary API.

15. Google LLC and Alphabet Inc. v European Commission — Google Android, Case T-604/18

Facts

The Google Android litigation concerned Google's agreements involving:

• Google Search;
• Chrome;
• Play Store;
• Android manufacturers;
• anti-fragmentation obligations;
• exclusivity arrangements.

The General Court treated Google's conduct within the broader context of a mobile ecosystem and exclusionary effects.

Principle

Dominance can be exercised across interconnected technological markets through contractual restrictions and ecosystem strategies.

AI orchestration relevance

This is particularly important for AI ecosystems.

An AI platform could potentially use:

Operating system dominance → AI access → application distribution → user data → AI improvement

to reinforce its position across multiple layers.

The case therefore illustrates the ecosystem leverage theory.

16. Google Shopping — Case C-48/22 P

Facts

Google was found to have abused its dominant position in general search by favouring its own comparison-shopping service within search results.

Principle

A dominant platform's control over an important intermediary can create competition concerns where it systematically gives preferential treatment to its own downstream service.

Application to AI orchestration

The same conceptual problem may arise where an AI orchestrator determines which AI agent receives a user's request.

For example:

User request

↓

Dominant AI orchestrator

↓

Choice between 10 AI agents

↓

Proprietary AI always receives priority

This raises a potential AI self-preferencing theory.

The key issue would not simply be ownership of the AI model but control over the gateway through which users reach competing models.

17. Eturas, Case C-74/14

Facts

Eturas concerned an electronic travel-booking system through which information concerning maximum discounts was communicated to participating travel agencies.

The case examined whether use of a computerized system could contribute to a concerted practice.

Principle

Digital infrastructure does not remove traditional competition-law rules concerning coordination.

AI-orchestration application

Suppose an AI pricing orchestrator automatically communicates market-sensitive information to competing businesses and those businesses knowingly rely upon the system.

Potential issues include:

• algorithmic coordination;
• information exchange;
• concerted practices;
• coordinated pricing;
• common algorithmic signals.

Thus, AI orchestration can create both exclusionary and collusive competition concerns.

18. T-Mobile Netherlands, Case C-8/08

Facts

The case concerned information exchange and coordination among mobile telecommunications operators.

Principle

Exchange of strategically sensitive information can restrict competition where it reduces uncertainty concerning competitors' market conduct.

AI-orchestration application

AI systems can dramatically increase the speed and frequency of information exchange.

Potentially sensitive information includes:

• prices;
• discounts;
• inventory;
• customer demand;
• capacity;
• future pricing strategies.

If an orchestration platform systematically collects and distributes such information among competing firms, traditional information-exchange principles may become relevant.

19. Additional Relevant Authority: Google Android Auto

The Google Android Auto litigation is particularly relevant to modern API-based ecosystems because it concerns access to functionality within Google's Android ecosystem.

The case has contributed to the continuing development of the debate surrounding interoperability and the application of refusal-to-supply principles in digital ecosystems. Contemporary EU competition commentary identifies the Android Auto ruling as significant for the development of the essential-facility doctrine in digital markets.

20. Synthesis of the Case Law

21. Possible Theories of Abuse

An AI-orchestration platform with substantial market power could potentially face several theories of harm.

1. Self-preferencing

Giving the platform's own AI preferential routing.

2. Refusal to interoperate

Preventing competing agents from accessing important APIs.

3. Tying

Making API access conditional upon adoption of another product.

4. Margin squeeze

Charging downstream competitors for APIs at rates that make effective competition difficult.

5. Discriminatory access

Providing inferior functionality to competing AI services.

6. Data foreclosure

Using proprietary data to improve the incumbent's AI while restricting competitors' access.

7. Exclusive dealing

Preventing developers from using competing orchestration platforms.

8. Raising rivals' costs

Increasing technical or financial costs for competing AI providers.

9. Algorithmic coordination

Using a common AI infrastructure to facilitate coordination among competitors.

10. Ecosystem leveraging

Transferring dominance from one layer to another.

22. The AI Orchestration "Stack"

Competition analysis should increasingly examine the entire stack:

Layer 1 — Compute

Cloud infrastructure and specialized chips.

↓

Layer 2 — Foundation Models

LLMs and other AI models.

↓

Layer 3 — APIs

Model and application interfaces.

↓

Layer 4 — Orchestration

Agent selection, routing and task allocation.

↓

Layer 5 — Applications

Enterprise software, search, productivity and commerce.

↓

Layer 6 — Distribution

Operating systems, app stores and devices.

↓

Layer 7 — Users

Consumers and businesses.

The most significant competition issue arises where one undertaking controls several layers simultaneously.

23. Network Effects and Feedback Loops

AI orchestration creates potentially powerful feedback loops:

More users

↓

More queries

↓

More behavioural data

↓

Better AI routing

↓

Better AI performance

↓

More developers

↓

More applications

↓

More users

This can create dynamic entry barriers.

A new competitor may possess an equally sophisticated AI model but still struggle because it lacks:

• users;
• API access;
• distribution;
• proprietary data;
• developer relationships;
• application integrations.

24. DMA and Ex Ante Regulation

Traditional Article 102-style analysis generally focuses on:

Dominance + abusive conduct + effects

The Digital Markets Act introduces a more ex-ante approach for designated gatekeepers.

The European Commission's July 2026 measures concerning Google's Android ecosystem specifically seek to ensure that competing AI services obtain effective access to Android functionality comparable to Google's own AI services.

This is significant because it demonstrates a shift from waiting for a conventional exclusionary-abuse case toward technical interoperability obligations imposed in advance.

The Commission has also been examining interoperability and contractual conditions in cloud markets, particularly in connection with large cloud providers and AI-related ecosystems.

25. Remedies

Possible remedies include:

A. Interoperability

Require equal technical access to APIs.

B. Non-discrimination

Require the platform to treat its own AI and rival AI services under equivalent conditions.

C. Data portability

Allow users and businesses to move relevant data between providers.

D. API transparency

Require clear documentation concerning:

• access;
• pricing;
• rate limits;
• technical requirements.

E. Separation

In extreme cases, structural separation between infrastructure and downstream AI services could be considered.

F. Switching rights

Reduce technical and contractual barriers to changing providers.

G. Monitoring

Require independent monitoring of API access and discrimination.

H. Data-sharing remedies

Where legally justified, require access to specified categories of strategically important data subject to privacy and security safeguards.

26. Objective Justifications

An API provider may have legitimate reasons for restricting access.

These can include:

• cybersecurity;
• privacy;
• system stability;
• fraud prevention;
• intellectual-property protection;
• technical limitations;
• protection against malicious AI agents;
• safety requirements.

The critical competition-law question is therefore not simply:

"Was access restricted?"

but:

Was the restriction objectively justified, proportionate and applied consistently rather than being a disguised method of excluding competitors?

The EU's current Android AI interoperability framework itself recognizes that security and data-protection risks can legitimately be considered when providing access.

27. Special Problem of AI Agents

AI agents create a distinctive competition issue because they can act through APIs, rather than merely display information.

An AI agent may:

• purchase goods;
• book travel;
• send emails;
• make payments;
• access enterprise databases;
• operate applications;
• execute financial transactions.

Consequently, control over the agent-to-API interface can become commercially more important than traditional application distribution.

A dominant orchestrator could potentially determine:

Which agent gets to act?

That makes the orchestration layer a possible competitive gateway.

28. India-Specific Relevance

Under India's Competition Act, 2002, the principal provisions likely to become relevant include:

• Section 4 — abuse of dominant position;
• Section 3 — anti-competitive agreements;
• Section 5 — combinations;
• Section 19 — inquiry and factors for determining relevant market/dominance.

Indian digital-platform jurisprudence concerning Google demonstrates the CCI's willingness to examine:

• network effects;
• ecosystem advantages;
• switching barriers;
• platform dependence;
• vertical relationships;
• leveraging of dominance.

The Google Play Store litigation is particularly relevant because the CCI recognized the importance of indirect network effects, developer dependence and barriers to alternative app stores in analysing Google's position.

The subsequent Indian appellate litigation also illustrates the continuing importance of defining the relevant digital market and assessing conduct across interconnected platform layers.

29. Practical Competition-Law Test

A regulator examining AI-orchestration dominance could use the following analytical sequence:

1. Identify the orchestration layer

↓

2. Define the relevant market

↓

3. Determine whether the platform is dominant

↓

4. Identify the controlled API/data/infrastructure

↓

5. Determine whether rivals depend upon it

↓

6. Examine the conduct

• self-preferencing?
• refusal?
• tying?
• discrimination?
• excessive API pricing?
• data foreclosure?
• exclusivity?

↓

7. Examine actual or potential foreclosure

↓

8. Consider efficiencies and objective justification

↓

9. Assess proportionality

↓

10. Select behavioural or structural remedies

30. Conclusion

API economy + AI orchestration creates a new potential competition bottleneck: control over the mechanism that decides which AI service, model, agent, application or API receives a user's task.

The central legal concern is therefore evolving from traditional questions of:

"Who owns the platform?"

to:

"Who controls access, routing, data and interoperability across the AI ecosystem?"

The established jurisprudence of Microsoft, IMS Health, Bronner, Google Android, Google Shopping, Eturas, T-Mobile Netherlands and Google Android Auto supplies the principal doctrinal building blocks.

The most important emerging theories are AI self-preferencing, API foreclosure, discriminatory interoperability, data foreclosure, tying, ecosystem leveraging, margin squeeze, switching barriers and algorithmic coordination.

Modern regulatory developments reinforce this direction: the EU's 2026 Android measures expressly seek to ensure that competing AI services can access critical Android functionality on an effective basis, illustrating how interoperability is becoming a central competition-law issue in AI ecosystems.