Global Registry of Gatekeeper Algorithms

1. Introduction

A Global Registry of Gatekeeper Algorithms would be a regulatory system requiring economically powerful digital platforms—particularly firms controlling search engines, app stores, operating systems, online marketplaces, advertising exchanges, social networks, browsers, and other digital gateways—to register specified algorithms that materially influence access to markets.

The idea is not presently a single worldwide legal institution. Rather, it represents a possible global competition-law and digital-regulation architecture built from developments such as the EU Digital Markets Act (DMA), national digital-market regimes, algorithmic-transparency initiatives, and traditional abuse-of-dominance jurisprudence.

The regulatory rationale is straightforward:

If a platform is a gatekeeper, its algorithm can become an instrument through which the gatekeeper determines who is visible, who receives access, who pays more, who is recommended, and ultimately who can compete.

The EU's DMA already moves in this direction. It currently supervises seven designated gatekeepers and 23 core platform services, while Article 15 requires independently audited information concerning consumer-profiling techniques.

2. Meaning of a Gatekeeper Algorithm

A gatekeeper algorithm is an algorithm operated by a platform possessing substantial control over access between businesses and users where the algorithm can materially affect competitive conditions.

Examples include:

• search-ranking algorithms;
• app-store ranking and recommendation algorithms;
• marketplace ranking algorithms;
• advertising-auction algorithms;
• recommendation engines;
• default-selection algorithms;
• price-ranking algorithms;
• seller-risk scoring systems;
• content-distribution algorithms;
• interoperability and access-control algorithms;
• algorithmic commissions;
• personalised pricing systems;
• algorithmic delisting systems;
• identity and verification algorithms;
• API-access algorithms;
• data-access and data-portability algorithms.

The crucial feature is economic gatekeeping power, rather than simply the existence of an algorithm.

3. Why a Global Registry Is Needed

A. Algorithms can determine market access

A marketplace may contain millions of products, but the platform's ranking algorithm determines which products consumers actually see.

Thus:

Algorithmic ranking → visibility → consumer traffic → sales → market share

A technically neutral-looking ranking system can therefore have substantial competitive consequences.

B. Algorithms can facilitate self-preferencing

A vertically integrated platform may rank its own service above competitors.

For example:

Platform owns search engine + shopping service

→ algorithm determines ranking

→ platform's own shopping service receives superior placement

→ competing comparison-shopping services receive less traffic

→ competitors become weaker

→ platform ecosystem becomes more entrenched.

This is one of the central concerns underlying modern digital competition regulation. The European Commission has recently treated Google's preferential ranking of its own services as a DMA compliance problem.

4. What Should Be Registered?

A global registry should not necessarily require publication of source code.

Instead, it could require registration of the economically significant characteristics of the algorithm.

Proposed Registry Categories

5. Registry Versus Source-Code Disclosure

This distinction is extremely important.

A competition authority generally does not need the complete source code to determine whether an algorithm is anti-competitive.

Instead, it may require:

Tier 1 — Basic registration

• algorithm's purpose;
• affected service;
• responsible entity;
• date of deployment.

Tier 2 — Regulatory disclosure

• principal ranking factors;
• relevant datasets;
• decision rules;
• optimisation objectives;
• changes affecting competitors.

Tier 3 — Confidential audit

Regulators receive access to:

• source code;
• model architecture;
• training data;
• logs;
• experiments;
• A/B testing;
• internal documentation.

Tier 4 — Public transparency

The public receives only:

• general explanation;
• audit findings;
• material changes;
• regulatory determinations.

This protects legitimate trade secrets while permitting competition authorities to investigate algorithmic discrimination.

6. Competition-Law Theory

A registry would be particularly useful where algorithms affect:

1. Self-preferencing

The gatekeeper gives preferential treatment to its own services.

2. Discriminatory ranking

Competitors receive inferior ranking without objective justification.

3. Algorithmic exclusion

An algorithm systematically reduces the visibility or access of rivals.

4. Predatory optimisation

The algorithm is deliberately configured to sacrifice short-term profits to eliminate competitors.

5. Algorithmic tying

Access to one service is technically or algorithmically conditioned upon adoption of another.

6. Data foreclosure

Competitors cannot access data necessary to compete effectively.

7. Algorithmic collusion

Competitors use automated pricing systems that facilitate coordinated outcomes.

8. Exploitative personalisation

The platform uses extensive data to identify users' willingness to pay and selectively alter commercial conditions.

9. Interoperability discrimination

The platform's algorithms make interoperability easier for affiliated services than for rivals.

10. Algorithmic retaliation

A platform reduces the ranking, visibility, or access of businesses that attempt to bypass its ecosystem.

7. Relationship With the Digital Markets Act

The DMA provides an important foundation for such a registry.

It regulates designated gatekeepers through obligations concerning ranking, access, interoperability, data use, steering, profiling and other forms of platform conduct. Unlike traditional competition-law proceedings, the DMA generally does not require the Commission to prove dominance and anti-competitive effects in the conventional Article 102 TFEU manner for every prohibited practice.

The present European system already contains significant transparency infrastructure:

• gatekeeper designation;
• core-platform-service identification;
• compliance reports;
• consumer-profiling reports;
• regulatory investigations;
• information requests;
• audits;
• compliance monitoring.

Consequently, a global algorithm registry could be viewed as an extension of this transparency model from platforms to the specific technological systems through which gatekeeping power is exercised.

8. Six Major Case Laws and Their Relevance

Case 1 — Google Shopping

Google Search (Shopping) — Google v Commission

This is perhaps the most important foundation for a gatekeeper-algorithm registry.

The dispute concerned Google's treatment of its own comparison-shopping service in search results.

The competition concern was not merely that Google operated a search engine. It was that the design of its ranking and display mechanism could systematically advantage Google's own service over competing comparison-shopping services.

Legal significance

The case demonstrates that:

Algorithmic ranking can itself constitute an important component of exclusionary conduct.

A registry would therefore allow authorities to identify:

• ranking algorithms;
• ranking criteria;
• treatment of affiliated services;
• changes to ranking systems;
• traffic allocation;
• demotion mechanisms.

Registry lesson:
Algorithms capable of determining competitive visibility should receive heightened regulatory scrutiny.

9. Case 2 — Google Android

Google Android — Google and Alphabet v Commission

The Android litigation concerned Google's contractual and technological arrangements surrounding Android and its associated services.

The broader competition issue involved Google's ability to use its position in one digital layer to reinforce its position in related markets.

Relevance to algorithm registries

Android demonstrates that gatekeeping is often ecosystemic.

A registry therefore should not examine an algorithm in isolation.

It should map:

Operating system → app store → search → browser → advertising → data → recommendation

An apparently minor algorithmic restriction can become highly significant when combined with contractual and technical restrictions.

Registry principle

The registry should therefore record algorithmic dependencies across an ecosystem, rather than treating every algorithm as an independent product.

10. Case 3 — Amazon Marketplace

Amazon Marketplace — Bundeskartellamt proceedings

Amazon has faced competition scrutiny concerning the relationship between its marketplace infrastructure and independent sellers.

The underlying concern in marketplace systems is particularly important for algorithmic regulation because the platform can simultaneously act as:

1. marketplace operator;
2. seller;
3. data collector;
4. ranking provider;
5. advertising intermediary;
6. logistics provider.

This creates a structural conflict.

Amazon can potentially possess information about:

• seller performance;
• sales volumes;
• consumer behaviour;
• product conversion;
• advertising performance;
• prices;
• inventory.

Registry implication

A gatekeeper registry should therefore record:

Who controls the algorithm?

Whose data trains it?

Whose products are evaluated by it?

Does the platform compete with the businesses being ranked?

This is essential to identifying algorithmic self-preferencing.

11. Case 4 — Epic Games v Apple

Epic Games v Apple

The dispute concerning Apple's App Store illustrates the importance of algorithmic and technological control over digital distribution.

The App Store is not merely a catalogue. Apple determines:

• technical admission;
• discoverability;
• distribution;
• payment architecture;
• ranking;
• security requirements;
• developer access.

Competition significance

A platform controlling the principal route through which consumers obtain applications may possess gateway power.

The algorithmic layer can therefore reinforce the contractual layer.

For example:

Developer → App Store review → ranking → recommendation → consumer discovery → download

Control over any of these stages can influence competitive opportunities.

Registry principle

A global registry should therefore cover not only algorithms that directly rank products, but also algorithms governing:

• app approval;
• visibility;
• recommendation;
• search;
• payment routing;
• access restrictions.

12. Case 5 — United States v Google / Google Search

Google Search antitrust litigation

The United States' major Google search litigation demonstrates the importance of control over search distribution, defaults and the wider search ecosystem.

Although the legal issues extend beyond algorithmic ranking itself, the case illustrates a fundamental principle:

Control over the gateway through which users reach information can generate durable competitive advantages.

A global registry would help competition authorities distinguish between:

• legitimate search optimisation;
• quality improvements;
• technical experimentation;

and:

• exclusionary manipulation;
• discriminatory treatment;
• strategic degradation of competitors.

The registry would therefore operate as an evidentiary infrastructure, rather than automatically treating every algorithmic change as unlawful.

13. Case 6 — Apple v Pepper

Apple Inc. v Pepper

The U.S. Supreme Court's Apple App Store litigation is important because it recognised the significance of the App Store's intermediary role for competition and antitrust standing.

The underlying economic structure is particularly relevant to gatekeeper regulation:

Apple → App Store → developers → consumers

The platform simultaneously controls the distribution channel through which consumers purchase applications.

Registry implication

A global registry should identify algorithms and technical systems controlling:

• application discoverability;
• search results;
• recommendations;
• payment routing;
• commission calculations;
• access restrictions.

The central question becomes:

Does the platform's algorithm merely facilitate transactions, or does it determine the competitive conditions under which transactions occur?

14. Case 7 — Facebook Germany

Bundeskartellamt — Facebook/Meta Data-Combination Proceedings

The German competition authority's proceedings concerning Facebook's combination and use of user data are highly relevant to an algorithm registry.

The fundamental issue is that data and algorithmic power are complementary.

A platform possessing enormous quantities of user data can use those data to:

• personalise content;
• optimise advertising;
• predict behaviour;
• rank information;
• improve recommendation systems;
• strengthen network effects.

Registry implication

Registration should therefore cover not only the algorithm but also the categories of data feeding economically significant algorithms.

The regulatory object becomes:

Data → Model → Algorithm → Market outcome

rather than merely "software."

15. Case 8 — Epic Games v Google

The Epic Games litigation concerning Google's Play Store provides another important illustration of digital-platform gatekeeping.

The case demonstrates how app-store rules, payment systems, distribution arrangements and technical architecture can operate together.

For algorithmic regulation, this supports a broader conception of gatekeeper algorithms:

An algorithm does not need to rank competitors to exercise gatekeeping power.

An algorithm that determines access, eligibility, payment routing, visibility or technical compatibility may also have substantial competitive consequences.

16. A Global Registry Model

A workable global system could have five layers.

Layer 1 — Identification

Identify:

• gatekeeper;
• service;
• algorithm;
• affected market.

Layer 2 — Risk classification

Algorithms could be classified:

Low risk

Routine operational systems.

Medium risk

Systems affecting business users.

High risk

Systems controlling:

• ranking;
• access;
• pricing;
• recommendations;
• advertising;
• interoperability.

Systemic risk

Algorithms capable of materially affecting an entire digital ecosystem.

17. Algorithm Change Notification

One of the most important features should be a material-change obligation.

A gatekeeper should notify regulators before—or immediately after—material changes involving:

• ranking;
• search visibility;
• seller eligibility;
• recommendation;
• advertising auctions;
• commission calculation;
• data access;
• interoperability.

For example:

Algorithm version 4.2 → ranking factor changed from 10% to 25%.

The regulator could then investigate whether the modification disadvantages competitors.

This would transform competition enforcement from a purely reactive model into a partially continuous monitoring model.

18. Algorithmic Audit

Independent auditors should examine:

A. Discrimination

Are equivalent firms treated differently?

B. Self-preferencing

Does the algorithm systematically favour affiliated services?

C. Transparency

Can affected businesses understand material ranking criteria?

D. Stability

Were algorithms deliberately changed following competitive threats?

E. Data advantage

Does the gatekeeper exploit data unavailable to rivals?

F. Feedback loops

Does the algorithm itself create the market power that it subsequently uses?

Example:

High ranking → more users → more data → better algorithm → higher ranking

This creates a potentially self-reinforcing competitive loop.

19. Algorithmic Feedback Loops

This is particularly important for AI-based platforms.

Consider:

Platform A has 70% market share

↓

more users

↓

more behavioural data

↓

better recommendation model

↓

higher conversion

↓

more sellers join

↓

more transactions

↓

more data

↓

stronger algorithm

↓

higher market share.

A conventional market-share analysis may describe the result but fail to identify the technological mechanism producing it.

A registry would allow regulators to monitor the algorithmic feedback loop itself.

20. Confidentiality and Trade Secrets

A major objection would be:

"A registry would force companies to reveal their proprietary algorithms."

That need not be the case.

A proper system should distinguish:

Public information

• algorithm purpose;
• affected market;
• risk category;
• audit status;
• material regulatory findings.

Confidential regulator information

• source code;
• model weights;
• training data;
• proprietary parameters;
• security architecture.

Highly restricted information

• cryptographic keys;
• cybersecurity vulnerabilities;
• commercially sensitive technical secrets.

Therefore:

Transparency ≠ complete publication of source code.

21. Global Governance Structure

A possible system could involve:

International coordination

• OECD;
• UNCTAD;
• competition authorities;
• digital-market regulators;
• data-protection authorities.

Regional authorities

• European Commission;
• UK Competition and Markets Authority;
• U.S. Federal Trade Commission;
• U.S. Department of Justice;
• national competition authorities.

National regulators

Each jurisdiction could maintain its own registry while exchanging:

• algorithm identifiers;
• audit results;
• infringement findings;
• risk classifications;
• material-change notifications.

22. Cross-Border Algorithm Passport

A particularly advanced model would create an Algorithm Passport.

Each important gatekeeper algorithm would receive a unique identifier.

For example:

GATE-ALPHA-SEARCH-0047

The record could contain:

• owner;
• service;
• jurisdiction;
• deployment date;
• risk category;
• material modifications;
• audit history;
• regulatory investigations;
• infringement findings.

This would prevent a company from substantially changing an algorithm in one jurisdiction while claiming that regulators elsewhere are dealing with a different system.

23. Interaction With GDPR

The registry would need to coordinate with data-protection law.

An algorithm may simultaneously raise:

Competition concerns

• exclusion;
• self-preferencing;
• discrimination.

Data-protection concerns

• profiling;
• excessive data processing;
• unlawful combination of data.

Consumer-protection concerns

• manipulation;
• dark patterns;
• personalised pricing.

AI-governance concerns

• transparency;
• accountability;
• automated decision-making.

Therefore, a global registry should be inter-regulatory, rather than being controlled exclusively by competition authorities.

24. Relationship With the DMA's Profiling System

The existing DMA framework is especially relevant because Article 15 requires gatekeepers to provide independently audited reports concerning consumer-profiling techniques and to publish non-confidential summaries.

This demonstrates a practical regulatory model for the proposed registry:

Gatekeeper designation

↓

Algorithm identification

↓

Profiling/ranking disclosure

↓

Independent audit

↓

Regulatory monitoring

↓

Compliance determination

↓

Remedial action

The European Commission has also increasingly moved toward continuing regulatory dialogue and monitoring rather than treating DMA enforcement as a one-time event.

25. Remedies for Algorithmic Abuse

If an algorithm is found to violate competition rules, remedies could include:

1. Transparency remedy

Require disclosure of relevant ranking factors.

2. Non-discrimination remedy

Require equal treatment of competing businesses.

3. Algorithmic separation

Separate the algorithm serving the gatekeeper's own business from the algorithm serving competitors.

4. Data-access remedy

Provide competitors with specified categories of data.

5. Interoperability remedy

Allow competing services to interact with the platform.

6. Monitoring remedy

Require continuous independent auditing.

7. Algorithmic injunction

Prohibit specified optimisation objectives.

8. Structural remedy

In extreme cases, separate platform infrastructure from competing commercial services.

26. Difficult Legal Questions

A global registry would create several difficult questions.

Question 1 — What qualifies as a gatekeeper?

Market share alone may be insufficient.

Relevant factors could include:

• user base;
• business-user dependency;
• switching costs;
• network effects;
• data advantage;
• ecosystem integration.

Question 2 — How much transparency is enough?

Complete transparency could permit gaming of algorithms.

Too little transparency could make enforcement impossible.

Question 3 — Who audits the algorithm?

Potential models include:

• government auditors;
• accredited private auditors;
• independent technical laboratories;
• multi-regulator audit teams.

Question 4 — Which country's law applies?

A global platform may deploy one algorithm worldwide.

Therefore:

One algorithm → multiple jurisdictions → conflicting regulatory standards

will become increasingly important.

27. Algorithmic Regulatory Capture

There is also a risk that gatekeepers could influence the regulatory system itself.

Large platforms possess:

• highly specialised engineers;
• lawyers;
• economists;
• cybersecurity specialists;
• massive technical resources.

A small regulator may consequently become dependent upon information supplied by the regulated company.

The registry should therefore guarantee:

• independent technical expertise;
• regulator access to logs;
• independent audits;
• whistleblower mechanisms;
• secure regulator testing environments.

28. Artificial Intelligence and Future Gatekeeper Algorithms

The problem becomes substantially more complicated with AI agents.

Traditional algorithm:

Input → predetermined rule → output.

AI system:

Data → model → inference → recommendation → feedback → model optimisation.

An AI-based gatekeeper may therefore continuously modify its behaviour.

This raises a new question:

What exactly should be registered—the algorithm, the model, the objective function, or the entire optimisation system?

For advanced AI platforms, the registry may need to record:

• model version;
• training-data categories;
• optimisation objectives;
• reinforcement signals;
• safety constraints;
• ranking objectives;
• automated experimentation;
• deployment environment;
• material behavioural changes.

29. Global Competition-Law Significance

The importance of a Global Registry of Gatekeeper Algorithms is that it changes the regulatory focus from:

"Is the company dominant?"

to:

"What technological mechanism allows the company to exercise gatekeeping power?"

This is a major conceptual shift.

Traditional competition law often examines:

Market → dominance → conduct → effects

Algorithmic regulation can additionally examine:

Technology → data → algorithm → access → ranking → market effect

The two approaches should complement rather than replace each other.

30. Key Case-Law Principles

31. Present European Direction

The European framework is increasingly close to the conceptual architecture required for such a registry.

As of the current framework, seven gatekeepers—Alphabet, Amazon, Apple, Booking, ByteDance, Meta and Microsoft—are subject to DMA obligations concerning 23 core platform services.

The Commission has also developed dedicated systems for compliance reports and consumer-profiling reports. In 2026, it additionally imposed major DMA penalties on Google concerning self-preferencing in Search and restrictions on steering users toward alternative purchasing channels.

These developments demonstrate why algorithmic transparency is moving from an academic proposal toward an important component of digital-market enforcement.

32. Conclusion

A Global Registry of Gatekeeper Algorithms would constitute a significant evolution in competition regulation.

Its central objective would not be to prohibit algorithms or force companies to disclose all source code. Instead, it would create an institutional mechanism for identifying which algorithms exercise economically significant gateway power, how they operate at a regulatory level, what data they use, how they change, and whether they systematically disadvantage competitors.

The strongest model would therefore be:

Gatekeeper designation + algorithm registration + confidential technical audit + public transparency + continuous monitoring + cross-border cooperation.

The most important legal lesson from Google Shopping, Google Android, Amazon marketplace proceedings, Apple App Store litigation, Facebook/Meta data cases and related digital-platform jurisprudence is that market power increasingly resides not merely in the platform itself, but in the technical systems through which the platform allocates visibility, access, data and transactions.

A global registry would make those systems identifiable, auditable and legally accountable, while preserving legitimate trade-secret protection. It could consequently become a foundational institution for the next generation of competition law, particularly as AI-powered gatekeepers begin making increasingly autonomous decisions about ranking, pricing, recommendations, access and market participation.