Competition Law and Adaptive Competition Institutions for AI Economies

1. Introduction

Adaptive competition institutions for AI economies refers to the development of competition authorities, regulatory mechanisms, investigative tools, and institutional structures that can respond to markets shaped by artificial intelligence.

Traditional competition institutions were largely designed around relatively stable markets where products, firms, prices, and market boundaries could be identified with reasonable clarity. AI economies are different because competition can depend on data, computing power, foundation models, cloud infrastructure, algorithms, talent, ecosystems, APIs, distribution channels, and strategic partnerships.

The OECD's recent work identifies AI competition concerns across several layers of the value chain, including foundation models, data, computing infrastructure, deployment, and downstream markets. It also emphasizes that competition authorities may need stronger technical expertise and continuous monitoring rather than relying only on traditional enforcement after harm occurs.

Therefore, an adaptive institution does not mean abandoning conventional competition law. It means updating the institutional capacity and methods used to apply competition law to rapidly changing AI markets.

2. Meaning of Adaptive Competition Institutions

An adaptive competition institution is a competition authority or institutional framework capable of:

1. continuously monitoring market developments;
2. acquiring technical AI expertise;
3. detecting new theories of competitive harm;
4. adapting merger-review techniques;
5. examining data and compute access;
6. investigating algorithmic conduct;
7. cooperating with other regulators;
8. using market studies and information-gathering powers;
9. responding to changing market structures; and
10. intervening without unnecessarily preventing beneficial innovation.

The central idea is:

Static markets can often be regulated through relatively static institutional methods; rapidly evolving AI markets require institutions capable of continuous learning and adjustment.

3. Why AI Economies Require Institutional Adaptation

A. AI markets are multilayered

The AI economy is not one market.

A simplified AI value chain can be represented as:

Chips → Computing infrastructure → Cloud → Data → Foundation models → APIs → AI applications → Consumers/businesses

A firm may have market power at one layer and use that position to strengthen its position at another.

For example:

• a cloud provider may also invest in an AI developer;
• a chip company may influence access to essential computing resources;
• a platform may control distribution of AI applications;
• an incumbent may possess valuable proprietary data;
• an AI developer may control an important model or API.

The OECD has specifically identified vertical integration, concentration, strategic partnerships and potential foreclosure across AI infrastructure as issues requiring competition authorities to monitor several interconnected layers.

4. Traditional Competition Institutions and Their Limitations

Traditional competition authorities normally rely on:

• merger control;
• abuse-of-dominance rules;
• cartel enforcement;
• market definition;
• economic analysis;
• consumer effects;
• investigations;
• remedies.

These tools remain relevant.

However, AI creates several additional institutional problems.

1. Rapid technological change

An investigation may take years while AI technology changes within months.

2. Difficult market definition

A single AI system may compete simultaneously with:

• another foundation model;
• traditional software;
• search engines;
• productivity applications;
• human labour;
• specialized software.

3. Non-price competition

AI services may be free or subsidized.

Competition may instead occur through:

• quality;
• accuracy;
• latency;
• model capabilities;
• privacy;
• interoperability;
• ecosystem access.

4. Data dependence

Data may be an important competitive input even where it is not directly sold.

5. Compute dependence

Advanced AI development requires substantial computing resources, making access to GPUs, data centres and cloud services potentially important.

6. Algorithmic decision-making

Pricing, recommendations, advertising and allocation may increasingly be performed by algorithms.

7. AI partnerships

A conventional acquisition may not be the only mechanism through which control or influence is obtained.

Strategic partnerships, investments, exclusivity arrangements and long-term compute agreements can also affect competitive conditions.

The FTC's study of major cloud/AI partnerships, for example, examined possible effects involving computing access, switching costs, technical and business information, and contractual arrangements.

5. Core Elements of an Adaptive Competition Institution

A. Permanent AI market monitoring

Competition authorities should not wait until a complaint arrives.

They can establish dedicated monitoring systems covering:

• foundation models;
• cloud computing;
• AI chips;
• data markets;
• AI APIs;
• AI application stores;
• model distribution;
• AI-related mergers;
• strategic investments;
• interoperability;
• switching costs.

This is particularly important because the AI market is developing quickly and its competitive effects remain uncertain. The OECD's 2026 work describes the AI environment as dynamic but uneven, with concentration concerns particularly relevant in some layers such as hardware and data.

6. Specialized AI Competition Units

A modern competition authority may need a multidisciplinary team consisting of:

• competition lawyers;
• economists;
• data scientists;
• machine-learning specialists;
• cloud engineers;
• cybersecurity specialists;
• software engineers;
• digital-market investigators.

This is important because a competition authority cannot properly evaluate an AI market if it cannot understand the technology underlying the alleged conduct.

The OECD has specifically highlighted the need for competition authorities to develop technical expertise across the AI compute stack.

7. Adaptive Merger Control

AI creates a particular problem for merger institutions.

A small AI company may have:

• valuable engineers;
• proprietary data;
• a promising model;
• a specialized technology;
• strategic importance to future competition.

Its current revenue may therefore underestimate its competitive importance.

Adaptive merger institutions should examine:

Acquisitions

• Is the target a potential future competitor?
• Does the acquirer obtain important AI talent?
• Does the transaction provide access to important data?
• Does it strengthen control over compute?
• Does it reinforce an existing ecosystem?

Minority investments

Even without complete ownership, an investment may create:

• influence;
• information access;
• exclusivity;
• dependence;
• switching costs.

Partnerships

Authorities may examine whether partnerships effectively produce:

• foreclosure;
• exclusivity;
• discriminatory access;
• dependency;
• reduced interoperability.

8. Case Law 1 — United States v. Microsoft Corp. (2001)

Background

Microsoft was accused of using its dominant position in PC operating systems to restrict competition from web browsers and competing technologies.

Importance

The case demonstrated that competition concerns can arise when a dominant firm uses control over one technological layer to influence competition in another.

Relevance to AI

The same institutional question can arise in AI ecosystems:

Can control over one layer of technology be used to disadvantage competitors at another layer?

For example:

Cloud → AI models → applications

An adaptive competition authority should therefore examine vertical relationships rather than treating each AI layer as completely independent.

Institutional lesson

Competition authorities need expertise capable of examining technology ecosystems and vertical relationships.

9. Case Law 2 — Google Search (Shopping), European Commission / General Court

The Google Shopping proceedings concerned Google's treatment of comparison-shopping services within its search ecosystem.

The broader significance is the examination of how a powerful digital platform can affect competition through the design and operation of its platform.

AI relevance

AI systems increasingly determine:

• search results;
• recommendations;
• rankings;
• visibility;
• access to consumers.

An AI platform could potentially influence which competing services receive exposure.

Institutional lesson

Competition authorities need the ability to investigate:

• ranking systems;
• algorithms;
• platform design;
• self-preferencing;
• discriminatory visibility.

This requires technical investigators rather than purely traditional legal analysis.

10. Case Law 3 — Google Android, European Commission

The Google Android case concerned several practices involving Google's Android ecosystem, including restrictions affecting competing services and applications.

Competition significance

The case illustrated how an ecosystem owner can use contractual and technical arrangements to strengthen its position across connected markets.

AI relevance

AI ecosystems may similarly combine:

• operating systems;
• cloud services;
• AI assistants;
• application stores;
• search;
• advertising;
• model distribution.

A dominant company controlling several layers may have opportunities to favor its own AI services.

Institutional lesson

Adaptive authorities must examine ecosystem effects, rather than looking at each contractual restriction in isolation.

11. Case Law 4 — United States v. Google LLC, Search and Advertising Litigation

The United States' antitrust litigation involving Google has focused on alleged exclusionary practices relating to search distribution and digital advertising.

AI relevance

The institutional lesson is broader than the specific technology involved.

A company controlling a major distribution channel can potentially influence the development of adjacent technologies.

For AI, authorities may therefore examine:

• default placement;
• distribution agreements;
• browser integration;
• device integration;
• AI assistant defaults;
• access to user interfaces.

Institutional lesson

AI competition analysis should consider distribution power, not only model quality.

12. Case Law 5 — FTC v. Qualcomm

The Qualcomm litigation concerned licensing practices and competition in cellular technology.

The case is important because it demonstrates the complexity of competition disputes involving:

• intellectual property;
• technology standards;
• licensing;
• component markets;
• vertical relationships.

AI relevance

AI markets also depend heavily on intellectual property and technical standards.

Examples include:

• model technology;
• semiconductor designs;
• AI accelerators;
• software frameworks;
• APIs;
• interoperability standards.

Institutional lesson

AI competition authorities need expertise at the intersection of:

Competition Law + Intellectual Property + Technology Standards.

13. Case Law 6 — Ohio v. American Express (2018)

The U.S. Supreme Court considered competition issues involving a two-sided transaction platform.

The case is particularly important for digital-market analysis because platform markets can connect different groups of users.

AI relevance

Many AI platforms are also multi-sided or ecosystem-based.

For example:

AI platform → developers + businesses + consumers

A competition authority may therefore need to understand effects on multiple groups rather than analysing only one side.

Institutional lesson

AI competition institutions need sophisticated economic methods for analysing multi-sided platforms and indirect network effects.

14. Case Law 7 — United States v. Apple / Epic Games v. Apple

The Apple-related competition litigation illustrates concerns surrounding:

• app distribution;
• platform control;
• commissions;
• payment systems;
• access restrictions;
• interoperability.

AI relevance

Similar issues may arise if dominant platforms control access to:

• AI assistants;
• model marketplaces;
• AI applications;
• APIs;
• AI agents;
• developer tools.

Institutional lesson

Competition institutions should examine whether platform rules that appear technical or contractual create significant barriers for AI competitors.

15. Case Law 8 — Microsoft Corp. v. Commission (T-201/04)

The European Microsoft litigation concerning interoperability and tying is especially relevant to institutional adaptation.

The case involved Microsoft's position in operating systems and its relationship with adjacent software markets.

AI relevance

Interoperability is becoming increasingly important in AI.

Examples include:

• model-to-model interoperability;
• data portability;
• API compatibility;
• cloud switching;
• AI-agent interoperability;
• application integration.

Institutional lesson

An adaptive institution must be technically capable of determining when interoperability restrictions create meaningful competitive disadvantages.

16. Algorithmic Competition and Adaptive Institutions

AI can change the way firms compete.

Algorithms may determine:

• prices;
• advertising;
• product rankings;
• discounts;
• inventory allocation;
• recommendations.

This creates several competition questions.

Algorithmic coordination

If competing firms use similar automated systems, authorities may need to determine whether the systems facilitate coordination.

Algorithmic discrimination

Algorithms could potentially discriminate against competitors by:

• lowering their visibility;
• restricting access;
• changing rankings;
• imposing different terms.

Algorithmic self-preferencing

A platform's AI system may favour its own products or services.

Institutional response

Competition agencies may need:

• algorithmic auditing;
• technical evidence;
• source-code or system-access powers where legally authorized;
• data analysis;
• expert testimony;
• continuous market monitoring.

The OECD's 2025 work on algorithmic pricing identifies AI-enabled algorithmic pricing as an emerging competition-policy issue across G7 jurisdictions.

17. Data as a Competitive Input

AI models require data.

Therefore, competition institutions may need to examine whether access to important datasets creates competitive advantages.

Potential concerns include:

Data foreclosure

A dominant firm may prevent rivals from obtaining important data.

Exclusive data agreements

Contracts may restrict competitors' access to valuable information.

Data combination

A firm may combine data from multiple markets, strengthening its position across an ecosystem.

Data portability

Users may face difficulty moving their data to competing AI services.

Institutional adaptation

Competition authorities may require:

• data-market studies;
• technical data analysis;
• interoperability investigations;
• assessment of data substitutability;
• examination of exclusive data agreements.

The OECD has identified access to quality data and computing power as potential barriers within the AI value chain.

18. Compute as a Competition Issue

Advanced AI depends on computing infrastructure.

Relevant inputs can include:

• GPUs;
• AI accelerators;
• cloud computing;
• data centres;
• networking;
• electricity;
• storage.

If access to these inputs becomes concentrated, competition authorities may need to consider whether competitors can realistically obtain comparable resources.

The OECD has described AI compute infrastructure as a multilayered supply chain involving chip design, manufacturing, servers, data centres, cloud services, networking and energy infrastructure.

19. Interoperability Institutions

Adaptive competition institutions may need to cooperate with technical regulators on interoperability.

Potential areas include:

Interoperability can reduce switching costs and make it easier for smaller competitors to enter or expand.

20. Competition Sandboxes

An adaptive institution may use controlled competition-policy sandboxes.

These can allow authorities to study:

• new AI business models;
• algorithmic pricing;
• model marketplaces;
• AI-agent markets;
• data-sharing systems;
• interoperability mechanisms.

The purpose is not to exempt firms from competition law but to help regulators understand emerging technologies before large-scale competitive problems develop.

21. Market Studies and Information-Gathering Powers

Traditional enforcement is reactive.

Market studies are more preventive.

An authority can study:

"How is the AI market developing, and where could competition problems emerge?"

rather than waiting for:

"Which company has already violated competition law?"

The FTC's investigation into large AI partnerships illustrates this institutional model: information-gathering powers were used to understand partnership structures and their potential competition implications.

22. International Cooperation

AI markets are global.

An AI developer may operate:

• model development in one country;
• cloud infrastructure in another;
• data centres in several countries;
• customers worldwide.

Therefore, competition authorities increasingly need cooperation through:

• information exchange;
• coordinated market studies;
• common analytical methods;
• technical workshops;
• joint research;
• enforcement coordination.

The OECD has also highlighted international cooperation as a means of allowing authorities to maintain knowledge and expertise as AI develops.

23. Adaptive Remedies

Traditional remedies may sometimes be too rigid for AI markets.

Possible remedies include:

Structural remedies

• divestiture;
• separation of businesses.

Behavioural remedies

• non-discrimination;
• access obligations;
• restrictions on exclusivity.

Technical remedies

• interoperability;
• API access;
• data portability.

Monitoring remedies

• independent compliance monitoring;
• periodic reporting;
• technical audits.

Time-limited remedies

Because technology changes quickly, some remedies may require periodic review.

24. Institutional Coordination

AI competition cannot always be handled by competition authorities alone.

Relevant institutions may include:

• competition authorities;
• AI regulators;
• data-protection authorities;
• telecommunications regulators;
• consumer-protection agencies;
• intellectual-property authorities;
• financial regulators;
• cybersecurity authorities.

This creates a multi-regulator governance model.

The EU's Digital Markets Act experience illustrates increasing institutional coordination around digital markets, while recent EU work has specifically examined the interaction between AI and digital-market regulation.

25. Adaptive Competition Institutions in India

In India, the Competition Commission of India (CCI) can potentially play an important role in AI-related competition questions through the existing competition-law framework.

Relevant areas include:

• abuse of dominant position;
• anti-competitive agreements;
• combinations;
• digital-platform conduct;
• access to important inputs;
• discriminatory conditions;
• tying and bundling;
• refusal to deal;
• exclusive arrangements.

AI markets may therefore be examined through existing competition principles while institutional capacity develops around:

• AI economics;
• data analysis;
• algorithms;
• cloud infrastructure;
• foundation models;
• digital ecosystems.

26. Adaptive Competition Institutions and Dominance

AI may produce several forms of dominance.

Data dominance

A firm controls unusually valuable datasets.

Compute dominance

A firm controls important computing infrastructure.

Model dominance

A firm controls an important foundation model.

Distribution dominance

A firm controls access to users.

Ecosystem dominance

A firm controls several interconnected AI layers.

An adaptive authority must therefore ask:

Where does competitive power actually come from?

rather than assuming that market share alone explains AI market power.

27. Adaptive Merger Assessment

AI merger review should potentially consider:

Current competition

Who competes with the parties today?

Potential competition

Who could compete with them tomorrow?

Innovation competition

Would the target have developed an important technology independently?

Input competition

Does the transaction affect access to:

• data;
• chips;
• cloud;
• talent?

Ecosystem competition

Does the transaction strengthen an existing ecosystem?

Cross-market effects

Could power from one market be transferred into another?

This is particularly important because AI start-ups can have relatively small current revenues while possessing strategically important technology or talent. Recent OECD evidence notes substantial AI start-up activity alongside frequent acquisition by large incumbents.

28. Institutional Learning

The word adaptive also means that institutions themselves must learn.

A competition authority should periodically review:

• whether its market-definition methods remain appropriate;
• whether existing merger thresholds capture important transactions;
• whether investigators possess sufficient technical expertise;
• whether remedies remain effective;
• whether AI developments have changed competitive conditions.

This resembles a continuous feedback cycle:

Monitor → Investigate → Intervene → Measure → Learn → Adapt

rather than:

Complaint → Investigation → Decision → End

29. Key Institutional Principles

An effective adaptive competition institution for AI economies can be based on ten principles:

1. Technological expertise
2. Continuous market monitoring
3. Flexible merger review
4. Data and compute analysis
5. Algorithmic investigation capability
6. Interoperability assessment
7. International cooperation
8. Cross-regulator coordination
9. Adaptive remedies
10. Periodic institutional review

30. Challenges of Adaptive Competition Institutions

Adaptive governance also creates risks.

A. Over-enforcement

Authorities could intervene before competitive harm is sufficiently established.

B. Under-enforcement

Authorities could wait too long while market power becomes entrenched.

C. Technical uncertainty

AI systems can be difficult to understand even for experts.

D. Regulatory overlap

Multiple regulators may investigate the same conduct.

E. Innovation concerns

Excessive intervention could potentially discourage investment or experimentation.

F. Institutional resource gaps

AI markets may develop faster than government expertise.

The OECD's 2025 competition trends work notes that competition authorities' resources and staffing have been increasing, while also showing substantial differences among jurisdictions.

31. Difference Between Traditional and Adaptive Competition Institutions

32. Overall Legal Framework

Adaptive competition institutions do not necessarily require an entirely new competition law.

A more practical approach can combine:

Existing competition law

• cartels;
• abuse of dominance;
• merger control;
• vertical restraints.

Digital-market regulation

• interoperability;
• platform obligations;
• access rules.

Institutional mechanisms

• market studies;
• technical investigations;
• expert teams;
• international cooperation.

AI-specific monitoring

• foundation models;
• compute;
• data;
• AI partnerships;
• algorithmic systems.

This creates a layered governance structure.

33. Conclusion

Competition Law and Adaptive Competition Institutions for AI Economies concerns the ability of competition authorities to evolve alongside AI-driven markets.

The central institutional challenge is that AI competition is not confined to a single product market. Competitive power can arise from a combination of:

Data + Compute + Models + Cloud + Talent + Distribution + Ecosystems + Algorithms.

The major case-law lessons from Microsoft, Google Shopping, Google Android, Google search-related litigation, Qualcomm, American Express, Apple/Epic and Microsoft interoperability litigation demonstrate why competition institutions increasingly need to understand technology ecosystems, platform power, interoperability, vertical integration and multi-sided markets.

The modern institutional model can therefore be expressed as:

Traditional competition law + AI technical expertise + continuous market monitoring + flexible merger control + data/compute analysis + algorithmic investigation + interoperability + international cooperation + adaptive remedies.

The most important point is that adaptation should occur at the institutional level without abandoning established principles of evidence, due process, economic analysis and proportionality. Current OECD work similarly emphasizes that AI's competitive effects remain highly context-dependent and that authorities need a combination of enforcement, advocacy, monitoring and technical expertise.