Competition Law and AI: Algorithmic Collusion, Market Power, and Merger Control

Algorithmic Collusion: Can AI Systems Conspire?

Traditional antitrust law prohibits agreements between competitors to fix prices or allocate markets (Article 101 TFEU, Sherman Act Section 1). Algorithmic collusion presents a challenge: AI pricing algorithms can converge on supra-competitive prices without explicit human communication. Competition scholars identify four scenarios of increasing concern: (1) algorithms implementing a human-agreed cartel (clearly illegal); (2) hub-and-spoke collusion through a shared algorithm or platform (illegal under existing law); (3) predictable agent behavior where algorithms learn to maintain high prices through repeated interaction (legally uncertain); (4) autonomous machine collusion without human design or awareness (legally uncertain). The European Commission's 2024 report on AI and competition acknowledges that scenarios 3 and 4 test the boundaries of existing prohibition frameworks, which require an "agreement" or "concerted practice."

Market Power Through Data and AI Capabilities

AI can entrench market dominance through feedback loops: dominant firms collect more data, train better models, attract more users, and collect even more data. This dynamic is recognized in the DMA's designation of gatekeepers and in abuse of dominance cases under Article 102 TFEU. The Bundeskartellamt's Meta/Facebook decision (2019, upheld by BGH 2020) established that data collection practices can constitute abuse of dominance. The European Commission's investigation of Microsoft/OpenAI (2024) examined whether investment structures in AI partnerships create de facto control without triggering merger notification thresholds.

Merger Control and AI Acquisitions

Competition authorities are adapting merger review to capture AI-related acquisitions that may not meet traditional turnover thresholds. The EU Merger Regulation allows the Commission to examine referrals from member states under Article 22, even for transactions below national thresholds, as applied in the Illumina/GRAIL case. Germany's Section 39a GWB allows the Bundeskartellamt to require notification of acquisitions in digital sectors where the target's turnover is low but its competitive significance is high. The UK CMA has asserted jurisdiction over AI partnerships and investment arrangements that may constitute relevant merger situations under the Enterprise Act 2002.

AI-Specific Competition Guidance

Several authorities have published AI competition guidance. The UK CMA's AI Foundation Models report (2023) identified access to compute, data, and technical talent as barriers to entry. The FTC's testimony before the US Senate (2023-2024) highlighted concerns about concentrated control of foundational AI infrastructure. The European Commission's DMA implementation treats AI-powered services within gatekeeper platforms as subject to obligations including data portability (Article 6(9)) and interoperability (Article 6(7)).

Compliance Strategies for AI Companies

Organizations using or developing AI should: (1) review pricing algorithms for features that could facilitate tacit coordination, particularly in oligopolistic markets; (2) assess whether data collection practices could constitute abuse if the organization holds a dominant position; (3) conduct self-assessment for DMA gatekeeper designation if operating core platform services with AI components; (4) evaluate AI-related acquisitions against both traditional turnover thresholds and newer jurisdictional hooks such as Article 22 referrals; and (5) maintain documentation of procompetitive justifications for data-sharing and algorithm design decisions.

This article is for informational purposes only and does not constitute legal advice. Regulatory requirements change frequently — verify current rules with official sources. Built by Sawai Gyoseishoshi Office, Hiroshima, Japan.