Competition Law and Research Automation Market Power

1. Introduction

Research automation refers to the use of software, artificial intelligence, algorithms, databases, automated laboratory systems, computational research tools, automated literature discovery, scientific-data platforms, simulation systems, and AI-based research assistants to perform or accelerate research and development.

Competition concerns arise when a research-automation provider obtains market power through control over essential research data, proprietary datasets, algorithms, APIs, scientific databases, interoperability standards, cloud infrastructure, laboratory platforms, or user-generated research information.

The central competition-law question is not whether automation itself is harmful. Rather, it is whether a firm with substantial market power uses control over a research-automation ecosystem to exclude rivals, raise rivals' costs, restrict interoperability, discriminate in access, tie complementary services, exploit data advantages, or prevent customers from switching.

A useful framework is:

Research automation + proprietary data + network effects + switching costs + interoperability restrictions = potential market-power concerns.

2. Meaning of Research Automation Market Power

Market power is the ability of an undertaking to behave to an appreciable extent independently of competitors, customers, or suppliers.

In research automation, market power may arise from several sources:

A. Proprietary research databases

A company may control:

• scientific publications;
• citation databases;
• clinical-trial information;
• patent datasets;
• laboratory results;
• research metadata;
• scientific benchmarks;
• training datasets.

If competitors cannot realistically reproduce the database, access restrictions can create significant competitive advantages.

B. Algorithmic superiority

An automated research platform may have algorithms that:

• identify relevant research;
• rank scientific literature;
• predict experimental outcomes;
• recommend research directions;
• generate hypotheses;
• automate simulations;
• identify molecular candidates;
• perform statistical analysis.

The algorithm can become an important competitive input.

C. Network effects

The value of an automated research platform may increase as more:

• researchers;
• universities;
• laboratories;
• pharmaceutical companies;
• publishers;
• data providers

use it.

This can create a feedback loop:

More users → more data → better automation → more users → more data.

D. Switching costs

Researchers may become dependent upon:

• proprietary data formats;
• APIs;
• workflows;
• laboratory integrations;
• stored research histories;
• customized models;
• proprietary computational environments.

Switching to another platform may therefore be expensive.

E. Ecosystem control

A dominant company may control several interconnected layers:

Data → computing → AI model → research software → laboratory automation → publication/distribution.

Competition problems become particularly significant where the same company controls multiple layers.

3. Relevant Markets

The relevant market must be defined carefully because "research automation" is extremely broad.

Possible relevant markets include:

1. automated scientific literature search;
2. AI research-assistant services;
3. scientific database services;
4. automated laboratory software;
5. computational drug-discovery platforms;
6. scientific cloud-computing services;
7. research-data management;
8. automated statistical-analysis software;
9. AI-based scientific discovery platforms;
10. laboratory information-management systems;
11. scientific simulation software;
12. research workflow automation.

A platform might therefore possess market power in one narrow market while facing substantial competition in another.

4. Sources of Market Power

4.1 Data advantages

A research-automation platform may accumulate an enormous proprietary dataset.

Competitors may face a significant barrier because:

Data accumulated through historical use can be difficult to replicate even when the underlying software is technically reproducible.

Competition authorities may therefore examine whether data advantages are:

• replicable;
• contestable;
• exclusive;
• portable;
• interoperable;
• protected by intellectual-property rights.

4.2 Proprietary APIs

An automated research platform may expose only limited APIs to competing services.

For example:

Dominant research platform → full API access to affiliated service

but:

Dominant research platform → restricted API access to independent competitor.

This can amount to discriminatory access where the API is competitively significant.

4.3 Interoperability restrictions

Research automation often depends on interoperability between:

• databases;
• cloud services;
• laboratory instruments;
• electronic laboratory notebooks;
• statistical software;
• AI models;
• research repositories.

A dominant undertaking could potentially weaken competitors by making interoperability unnecessarily difficult.

4.4 Tying and bundling

A dominant research-data provider might require customers purchasing:

scientific database access

also to purchase:

its proprietary AI research assistant.

Alternatively, laboratory-automation software might be bundled with proprietary cloud computing.

The relevant questions include:

• Are the products separate?
• Does the firm possess dominance in the tying market?
• Are customers effectively forced to accept the tied product?
• Does the practice foreclose competitors?
• Is there an objective justification?

5. Six Important Case Laws

There is not yet a large body of reported decisions specifically involving AI-powered research automation. Consequently, the most useful authorities come from cases involving databases, information infrastructure, digital platforms, interoperability, refusal of access, tying, and technology-driven market power.

Case 1: IMS Health GmbH & Co. OHG v NDC Health GmbH & Co. KG

Court: Court of Justice of the European Union
Year: 2004

Facts

IMS Health operated a system for pharmaceutical-sales data organized into geographic structures known as the "1860 brick structure."

Competitors experienced difficulty entering the market because pharmaceutical companies had become accustomed to using the established structure.

Legal issue

The case concerned whether refusal by a dominant undertaking to license an intellectual-property-protected system could constitute an abuse of dominance.

Principle

The Court established stringent conditions for compulsory licensing, including circumstances in which the refusal prevents the emergence of a new product for which there is consumer demand, lacks justification, and reserves a market to the dominant undertaking.

Relevance to research automation

The case is highly relevant where an automated research platform controls:

• a proprietary scientific classification system;
• research-data architecture;
• a unique database format;
• an essential scientific information structure.

A dominant research platform cannot automatically be required to license its technology merely because competitors would benefit. The IMS Health conditions demonstrate that compulsory access is an exceptional remedy.

6. Case 2: Bronner v Mediaprint

Case: Oscar Bronner GmbH & Co. KG v Mediaprint Zeitungs und Zeitschriftenverlag GmbH

Court: CJEU
Year: 1998

Facts

Mediaprint operated an extensive newspaper home-delivery network. Bronner sought access to that network.

Principle

The CJEU developed important conditions for treating a refusal to provide access to infrastructure as abusive.

The facility must generally be indispensable, meaning that there must be no actual or potential alternative that is technically, legally, or economically viable.

Application to research automation

Suppose a dominant scientific platform controls:

• the only practically usable automated research database;
• an irreplaceable laboratory-data interface;
• an indispensable scientific API.

A refusal to provide access may raise essential-facility/refusal-to-deal concerns.

However, mere usefulness is insufficient.

The competitor generally must demonstrate something much closer to indispensability.

7. Case 3: Microsoft Corp. v Commission

Court: General Court of the European Union
Year: 2007

Facts

Microsoft was found to have abused its dominant position through, among other conduct, restrictions concerning interoperability information and tying.

Importance

The interoperability component was particularly significant.

The Commission and EU courts examined whether Microsoft had withheld information necessary for competitors to achieve interoperability with Microsoft's dominant operating-system environment.

Research-automation relevance

The same principle can arise where a dominant research platform controls:

• APIs;
• data formats;
• interoperability protocols;
• authentication systems;
• research-workflow interfaces.

For example:

Dominant platform → proprietary research database → proprietary API → independent research automation tools

If access restrictions substantially impair interoperability, competition authorities may investigate whether the conduct forecloses competing systems.

8. Case 4: Google Shopping

Case: Google Search (Shopping)
Authority: European Commission / General Court litigation

Facts

Google was found to have treated its own comparison-shopping service more favourably in general search results than competing comparison-shopping services.

Competition principle

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

Research-automation application

Consider a dominant scientific-search platform operating:

1. a scientific search engine;
2. a research-automation assistant.

If the platform systematically gives its own research-automation service preferential visibility while disadvantaging competing research assistants, the conduct may raise self-preferencing concerns.

The relevant economic question is whether the platform is using dominance at one level to distort competition at another level.

9. Case 5: Slovak Telekom v Commission

Court: CJEU
Year: 2021

Facts

Slovak Telekom, part of the Deutsche Telekom group, was found to have engaged in conduct concerning access to its network infrastructure that restricted competition.

Importance

The case demonstrates how competition law approaches exclusionary conduct involving access to infrastructure controlled by a dominant undertaking.

Research-automation relevance

Research automation increasingly depends upon infrastructure such as:

• cloud computing;
• specialized computing capacity;
• data repositories;
• laboratory networks;
• scientific APIs.

Where infrastructure is controlled by a dominant undertaking, discriminatory or exclusionary access conditions may become an important competition issue.

The analysis must distinguish legitimate infrastructure management from conduct designed to exclude rivals.

10. Case 6: Google Android

Case: Google Android
Authority: European Commission / General Court

Issues

The case concerned Google's use of contractual arrangements involving Android devices, including restrictions associated with:

• Google Search;
• Google Play;
• browser distribution;
• mobile-app ecosystems.

Competition significance

The case illustrates how dominance can potentially be leveraged across interconnected products.

Research-automation application

An analogous problem could arise if a dominant scientific-computing provider conditions access to:

research cloud infrastructure

on the use of:

its own AI research assistant, database, laboratory software, or model marketplace.

This could create ecosystem leverage and make it harder for independent research-automation providers to compete.

11. Case 7: Magill

Cases: RTE and ITP v Commission — Magill
Court: CJEU
Year: 1991

Facts

Television broadcasters controlled copyright-protected programme listings. Magill sought to publish comprehensive television listings.

Principle

The case became a foundational authority concerning exceptional circumstances in which refusal to license intellectual property may constitute abuse of dominance.

Research-automation relevance

Scientific information providers may possess copyright or other intellectual-property rights over:

• datasets;
• structured research information;
• scientific metadata;
• databases.

The existence of IP rights does not automatically determine the competition-law question.

Where a dominant undertaking uses IP rights to control an important downstream information market, Magill-type principles may become relevant.

12. Case 8: Google Search (AdSense)

Authority: European Commission
Year: 2019

Facts

The Commission examined contractual restrictions used by Google concerning the placement of search advertisements on third-party websites.

Competition significance

The case demonstrates how contractual restrictions can protect a dominant position by limiting opportunities available to competing providers.

Research-automation analogy

A dominant research platform could potentially impose contractual provisions preventing universities, laboratories, or researchers from:

• using competing research assistants;
• exporting research data;
• integrating competing AI tools;
• simultaneously using rival databases.

Such provisions may increase switching costs and restrict multi-homing.

13. Research Automation and Refusal to Supply

A particularly important competition-law problem is:

When does refusal to provide automated research inputs become an abuse of dominance?

Potential inputs include:

• research datasets;
• APIs;
• metadata;
• model interfaces;
• laboratory protocols;
• scientific databases;
• computing infrastructure.

The analysis should normally examine:

Step 1 — Dominance

Is the provider dominant in a properly defined relevant market?

Step 2 — Controlled input

Does the undertaking control an input that competitors genuinely require?

Step 3 — Indispensability

Are there realistic alternatives?

Step 4 — Competitive foreclosure

Would denial of access substantially impair competition?

Step 5 — Justification

Does the undertaking have legitimate reasons for restricting access?

Step 6 — Remedy

Would access, licensing, interoperability, or non-discrimination obligations be proportionate?

14. Algorithmic Discrimination

Research-automation platforms may rank:

• papers;
• laboratories;
• researchers;
• datasets;
• research tools;
• experimental recommendations.

A dominant platform could potentially manipulate rankings to favor its own ecosystem.

For example:

Independent AI research tool

↓ ranking

Platform's own research assistant

↑ ranking

Such conduct may raise concerns similar to the broader problem of platform self-preferencing.

Competition authorities would need to establish that the ranking mechanism is capable of harming competition rather than merely reflecting legitimate relevance or quality criteria.

15. Data Lock-In

Data lock-in is particularly important.

A researcher might have years of:

• experimental records;
• annotations;
• workflow configurations;
• model fine-tuning;
• research histories;
• metadata

stored inside one platform.

If the platform does not provide meaningful data portability, the customer may face significant switching costs.

This can create:

Customer lock-in → reduced switching → reduced competitive pressure → stronger market power.

However, high switching costs alone do not establish an infringement. Their competitive significance depends upon the market structure and the conduct of the undertaking.

16. Network Effects in Research Automation

Network effects can reinforce market power.

Direct network effect

More researchers use the platform → the platform becomes more valuable to researchers.

Indirect network effect

More researchers → more developers create compatible tools → platform becomes more attractive.

Data network effect

More usage → more data → better algorithms → better results → more usage.

This third category is particularly important for AI-based research automation.

17. AI and Research Automation

AI may amplify existing competition concerns because AI systems often depend upon:

• enormous datasets;
• computing infrastructure;
• specialized models;
• user feedback;
• proprietary interfaces;
• distribution channels.

A large incumbent may therefore possess several mutually reinforcing advantages.

Potential competitive loop

Large user base

↓

More research interactions

↓

More proprietary usage data

↓

Improved AI system

↓

Better research automation

↓

Higher customer retention

↓

More users

This may create a powerful feedback mechanism.

18. Killer Acquisitions

Research automation markets may also raise merger-control concerns.

An established scientific-data company could acquire:

• an AI research startup;
• an automated laboratory company;
• a scientific-search engine;
• a specialized research-data provider.

The acquisition may be strategically important even when the target has relatively low current revenue.

Competition authorities may therefore examine:

• innovation competition;
• potential competition;
• data assets;
• pipeline products;
• nascent technologies;
• access to complementary datasets;
• future interoperability.

19. Bundling and Ecosystem Expansion

Suppose a dominant company controls:

Scientific database + cloud + AI model + laboratory software.

It could offer:

Database + AI assistant + cloud computing + laboratory automation

as a single package.

This can generate legitimate efficiencies, but competition authorities may investigate whether bundling:

• forecloses independent providers;
• raises rivals' costs;
• prevents multi-homing;
• increases switching costs;
• makes independent research tools commercially unviable.

20. Exclusive Dealing

Research automation platforms may seek exclusive agreements with:

• universities;
• hospitals;
• laboratories;
• pharmaceutical companies;
• research institutes;
• scientific publishers.

Exclusive arrangements may reduce the available customer base for competing platforms.

Their legal assessment generally depends upon factors such as:

• duration;
• market coverage;
• foreclosure percentage;
• market power;
• customer alternatives;
• efficiency justifications.

21. Interoperability as a Competition Remedy

Where competition problems arise from technological incompatibility, possible remedies may include:

1. API access

Competitors receive standardized access to interfaces.

2. Data portability

Users can export their research information.

3. Interoperability

Independent software can communicate with the dominant platform.

4. Non-discrimination

The dominant platform cannot provide materially better access to its own affiliated services.

5. Data separation

Sensitive competitive information may be separated between different business units.

6. Choice screens

Customers may be offered meaningful choices among competing research services.

22. Competition Law and Intellectual Property

Research automation frequently involves intellectual property.

The basic principle is:

Intellectual-property ownership does not automatically create competition-law immunity.

At the same time:

Competition law does not automatically require a dominant firm to license every proprietary technology.

The difficult middle ground is determining when IP control becomes a mechanism for exclusion.

The Magill, IMS Health and Microsoft lines of authority are therefore particularly useful.

23. Economic Evidence

Competition authorities may examine quantitative evidence such as:

• market shares;
• concentration ratios;
• HHI;
• switching rates;
• customer churn;
• API usage;
• data-access costs;
• price-cost margins;
• foreclosure percentages;
• customer acquisition costs;
• multi-homing rates;
• interoperability costs;
• entry barriers.

For research automation, conventional market-share analysis may be insufficient because services may be provided at zero monetary prices.

Data, attention, research time, and ecosystem dependence can therefore become important competitive variables.

24. Key Competition Concerns

25. Regulatory Challenges

Research automation creates several difficulties for competition authorities.

A. Rapid technological change

Market definitions can become obsolete quickly.

B. Zero-price services

Traditional price-based analysis may not capture competitive harm.

C. Data advantages

It can be difficult to determine whether a dataset is genuinely irreplaceable.

D. Algorithmic opacity

Authorities may need to understand how automated ranking or recommendation systems operate.

E. Innovation competition

A conduct that appears exclusionary in the short term might sometimes produce technological efficiencies.

F. Global markets

Research automation platforms may serve researchers worldwide, creating jurisdictional and regulatory complexity.

26. Indian Competition-Law Perspective

Under the Competition Act, 2002, research-automation market power can principally be examined through the framework of:

• Section 4 — abuse of dominant position;
• Section 3 — anti-competitive agreements;
• Sections 5 and 6 — combinations.

Potential Section 4 theories include:

• discriminatory access;
• unfair conditions;
• denial of market access;
• tying or bundling;
• leveraging dominance;
• exclusionary conduct.

Section 3 may become relevant where research platforms or suppliers coordinate on:

• prices;
• access conditions;
• licensing;
• customer allocation;
• data-sharing arrangements.

For mergers, competition authorities may examine whether an acquisition removes an important emerging competitor or combines complementary data and technological assets.

27. Practical Analytical Framework

A research-automation competition case can be analyzed through the following flow:

Identify technology

↓

Define relevant product/service market

↓

Define geographic market

↓

Measure market power

↓

Identify source of power

↓

Examine conduct

↓

Determine foreclosure effects

↓

Assess efficiencies and objective justification

↓

Assess effect on innovation

↓

Consider proportionate remedy

28. Conclusion

Research automation is likely to make data, interoperability, algorithms, computing infrastructure and ecosystem control increasingly important sources of market power.

The principal competition-law danger is not automation itself. The concern arises where a powerful research-automation provider uses control over a strategically important input or ecosystem to exclude competitors, discriminate against rival services, prevent interoperability, lock in customers, or extend dominance into adjacent research markets.

The most useful established authorities include Magill, IMS Health, Bronner, Microsoft, Google Shopping, Slovak Telekom, Google Android and Google AdSense. They do not all concern modern AI research automation directly; rather, they provide the established competition-law principles that can be applied to emerging research-automation markets.