Adaptive Antitrust Agency Enforcement and the Use of AI: Legal and Policy Implications
Abstract
Antitrust enforcement increasingly targets markets where pricing, product visibility and consumer interaction are mediated by adaptive AI systems. This Article argues that the core enforcement challenge is not complexity but continuous redesign: firms use algorithmic tools to alter the technical manifestation of harmful conduct while preserving supracompetitive prices, collusion or exclusionary effects. We distinguish detection from recognition, and show that enforcement effectiveness depends on how quickly agencies learn new patterns of conduct. When recognition is slow, firms can engage in persistent “cat-and-mouse” redesign, creating an arms race between agency and firm AI capabilities. We develop a framework in which AI investments by agencies reduce harm only if they cross a deterrence threshold; sub?threshold, partial adoption can worsen congestion and increase workload without diminishing aggregate harm. Although elaborated in an antitrust setting, the recognition?redesign dynamic generalizes to consumer protection domains such as phishing, robocalls and investment scams, where legality is clearer and adaptive evasion is pervasive. The Article concludes that frustration with case?by?case antitrust enforcement should not automatically justify broad ex ante regulation; instead, existing institutions must become more adaptive in their use of AI or risk declining relevance.
Read the full piece at SSRN.