AI Enforcement in Europe: How Algorithms Are Reshaping Antitrust in the EU and UK

The Michelin Moment

The clearest signal came from an unlikely source — earnings calls.

The Commission used AI tools to screen publicly available earnings calls for price signalling patterns. When the analysis flagged suspicious behaviour, inspectors arrived at Michelin’s door. The company challenged the raid, arguing the AI evidence was insufficient. The EU General Court disagreed.

That ruling matters enormously. It establishes that AI-detected patterns can justify formal investigation. If that precedent holds, every major company’s investor communications just became a compliance risk surface.

Spain, Portugal, and the Network Effect

Enforcement isn’t just centralising in Brussels. National competition authorities are building their own tools and sharing them across the European Competition Network.

Spain’s market-screening software is already live and being distributed to peers. Portugal is cross-referencing news sources and company registries to catch gun-jumping in M&A. The infrastructure for pan-European AI-assisted enforcement is assembling quietly, jurisdiction by jurisdiction.

Algorithmic Pricing: Efficiency or Collusion?

The CMA’s position is nuanced but firm. Algorithms can be genuinely pro-competitive — faster price adjustments, better resource allocation, more personalised services. But the same systems can facilitate coordination between competitors without anyone ever picking up the phone.

The CMA is already investigating hotel chains over potential algorithmic coordination through shared data analytics tools. No explicit agreement required. Just predictable, parallel behaviour enabled by third-party software.

That’s a significant legal frontier. Tacit collusion via algorithm is harder to prove — and harder to defend against.

Agentic AI: The Next Frontier

The CMA is paying close attention to agentic AI — systems that don’t just recommend actions but execute them autonomously. These agents can monitor markets, adjust pricing, interact with consumers, and make commercial decisions at machine speed.

The competitive implications are significant. When every major player deploys autonomous agents optimising against each other in real time, market dynamics can shift in ways that look a lot like coordination — even without intent.

The CMA has flagged this explicitly. Guidance is coming. Investigations will follow.

EU: A Framework on Top of a Framework

The EU AI Act becomes fully applicable in August 2026, layering a risk-based AI governance regime on top of existing competition tools. High-risk AI systems face transparency and oversight obligations. Some applications are simply prohibited.

Alongside this, the Digital Markets Act imposes interoperability and anti-tying obligations on designated platforms — with AI activities now explicitly in scope for transaction reporting. The draft EU Merger Guidelines published in April 2026 go further, bringing algorithmic pricing directly into merger assessment. Acqui-hires of AI talent are now a merger control consideration.

The regulatory surface area is expanding fast.

UK: New Powers, New Expectations

The Digital Markets, Competition and Consumers Act 2024 gives the CMA new investigatory muscle — including the ability to require firms with strategic market status to generate data and run tests. Algorithms are the obvious use case.

The CMA’s leniency policy now explicitly covers algorithmic information exchange between competitors. That’s a meaningful signal: regulators expect businesses to self-report AI-facilitated coordination, not just human-orchestrated cartels.

A post-Brexit UK-EU Competition Cooperation Agreement is also incoming — the first of its kind. Cross-border enforcement coordination is about to get structurally tighter.

Compliance Doesn’t Stop at the Human Layer

If your pricing engine, recommendation system, or procurement tool produces anti-competitive outcomes, the fact that no human made that specific decision is irrelevant. The obligation runs to the output, not the input.

Governance frameworks need to extend into the algorithmic layer — not just wrap around it.

Speed Is Now a Regulatory Variable

AI-assisted enforcement moves faster. The Commission can screen thousands of earnings calls in the time it once took to review a single document set. The CMA can identify bid-rigging patterns in public procurement at scale.

Businesses that rely on slow enforcement cycles as informal risk buffers should update their models. The detection window is shrinking.

Structural Factors Are Becoming Strategic

The CMA is increasingly focused on the infrastructure beneath AI markets — data portability, interoperability standards, digital identity. These aren’t abstract policy concerns. They determine who can compete, who gets locked in, and which market structures become entrenched.

Companies that understand these structural dynamics will navigate regulatory risk more effectively than those focused only on product-level compliance.