Source: Wolters Kluwer (Competition Blog)

All 27 EU national competition agencies, DG Competition, and many agencies worldwide now rely on computational tools to enforce competition law, with concrete examples including Spain’s BRAVA for bid-rigging screening, Brazil’s Cerebro for procurement-document analysis, France’s natural-language queries over case databases using retrieval-augmented-generation built on large language models, Greek analysis of email metadata from dawn raids, Chile’s monitoring of over 80,000 products for price anomalies, and Poland’s investigation of dark patterns via eye-tracking and neuromarketing methods. The practice has a name—computational antitrust—and it has shifted from experiment to routine within under a decade.

The article argues that the existing legal framework lags behind these practices in two main respects. First, collection: rules such as Regulation 1/2003 and the ECN+ Directive were drafted for document requests and inspections, not continuous large-scale machine-readable data flows, producing tension with proportionality limits, business secrecy, data protection, and privilege against self-incrimination. Second, processing: AI systems inherit training-data biases, and screening or model-driven selections reflect design and data choices outside external scrutiny, creating explainability problems for the duty to give reasons; the AI Act layers obligations but leaves open how it applies when enforcement agencies use AI against private parties. The post warns of annulment and legitimacy risks if doctrinal work does not follow, and describes ATLANTIS, a five-year ERC-funded research project at Vrije Universiteit Amsterdam focused on accuracy, fairness, and institutional arrangements; it announces recruitment of three fully funded PhD positions (legal and computational tracks) starting 1 December 2026, with applications due 15 September 2026 via VU vacancy 5706, and summarizes standard Dutch employment conditions, networked fieldwork opportunities, and ERC grant details (no. 101228709).