Case 9: United States v. Topkins (2015)

Court: United States District Court, Northern District of California.

The matter concerned an online seller who participated in an agreement to fix prices for posters sold through Amazon Marketplace, using pricing software to implement the agreed prices. The case resulted in a criminal antitrust prosecution.

Principle: Software does not shield businesses from liability where it implements a human agreement to fix prices.

Relevance to computational governance: Companies using repricing tools should assess who sets pricing rules, what information the software uses, whether competitors have agreed on pricing strategies, and whether the system implements any unlawful arrangement. This is a US criminal antitrust matter, not binding Indian precedent.

Case 10: United States v. David Topkins — algorithmic implementation of price-fixing (2015)

This is the same prosecution commonly cited as United States v. Topkins, rather than a separate independent precedent. It is mentioned here to clarify that the key legal lesson arises from one prosecution, not two distinct cases.

The case illustrates a core principle of computational antitrust governance: a business must examine both the commercial agreement and the software that implements it. A pricing algorithm can be a mechanism for carrying out unlawful coordination, even where employees do not manually change each price.

7. Practical framework for computational antitrust governance

An organisation can implement computational antitrust governance through the following controls:

Algorithm inventory: Record all systems used for pricing, ranking, bidding, allocation, recommendations, and market monitoring.

Legal risk assessment: Determine whether each system could facilitate coordination, exclusion, discriminatory treatment, or abuse of market power.

Data governance: Review data sources, access permissions, retention, and whether commercially sensitive competitor information is used.

Model validation: Test whether outputs or optimisation objectives could create anti-competitive effects, including through repeated price matching or restrictions on rivals.

Human accountability: Assign responsible business and technical owners; do not treat algorithmic decisions as outside corporate responsibility.

Monitoring and audit: Retain appropriate logs, investigate anomalies, and test systems after material changes.

Incident response: Provide a process to suspend or modify risky functionality, preserve evidence, seek legal advice, and implement corrective action.

Periodic reassessment: Review systems after changes in market conditions, model architecture, data inputs, or relevant law.

8. Challenges and limitations

Computational antitrust governance faces several difficulties:

Opacity: Complex models may be hard to explain or audit.

Rapid adaptation: Self-learning systems can change their behaviour after deployment.

False positives: Similar prices or parallel outputs do not necessarily establish an unlawful agreement.

Responsibility gaps: Businesses may incorrectly attribute decisions to software vendors or autonomous systems.

Data sensitivity: Investigations must be conducted with appropriate confidentiality, privacy, and security safeguards.

Legal uncertainty: Existing competition-law concepts may not map neatly onto every form of autonomous algorithmic coordination.

The OECD's work on algorithmic competition stresses that algorithms can improve efficiency and benefit consumers, but they can also facilitate anti-competitive outcomes. Governance should therefore be proportionate to the system's function, market context, and risks.

OECD

+2

9. Conclusion

Computational antitrust governance integrates competition law with algorithm design, data management, software auditing, market analysis, and corporate accountability. In India, Sections 3 and 4 of the Competition Act provide the central framework for assessing anti-competitive agreements and abuse of dominance, while Sections 5 and 6 are relevant to merger control.

The cases discussed demonstrate that software can implement unlawful price-fixing, digital platforms can create new forms of coordination risk, and market data must be evaluated carefully rather than treated as conclusive proof. Samir Agrawal, in particular, illustrates the importance of identifying the agreement or concerted conduct required by the applicable legal framework.

The core principle is that algorithms must remain subject to competition-law oversight. Organisations should understand how their systems make commercial decisions, monitor for anti-competitive effects, preserve meaningful human accountability, and correct identified risks without assuming that the use of technology either proves or excuses a legal violation.## 10. Conclusion

Computational antitrust governance integrates competition law with algorithm design, data management, software auditing, market analysis, and corporate accountability. In India, Sections 3 and 4 of the Competition Act provide the central framework for assessing anti-competitive agreements and abuse of dominance, while Sections 5 and 6 are relevant to merger control.

The cases discussed demonstrate that software can implement unlawful price-fixing, digital platforms can create new forms of coordination risk, and market data must be evaluated carefully rather than treated as conclusive proof. Samir Agrawal, in particular, illustrates the importance of identifying the agreement or concerted conduct required by the applicable legal framework.

The core principle is that algorithms must remain subject to competition-law oversight. Organisations should understand how their systems make commercial decisions, monitor for anti-competitive effects, preserve meaningful human accountability, and correct identified risks without assuming that the use of technology either proves or excuses a legal violation.