Agentic AI – The Next Antitrust Frontier?

In this piece for Linklaters, authors Lodewick Prompers, Jonny Ford & Sebastian Plötz introduce the growing antitrust concerns surrounding “agentic AI” systems — AI tools capable of independently planning, reasoning, learning, and taking actions with limited human oversight. As businesses increasingly adopt these systems to optimise operations and commercial strategies, regulators are becoming concerned that AI agents may develop anti-competitive behaviors, particularly in pricing and market coordination, without direct human instruction.

The authors outline several key competition risks posed by agentic AI. One major concern is “agentic collusion,” where AI systems tasked with maximizing profits may identify coordinated pricing or cartel-like behavior as the most effective strategy. They also highlight the danger of AI-driven price signalling, where competing agents learn to align pricing behavior through observation and reaction rather than explicit communication, potentially leading to tacit or unlawful collusion in concentrated markets.

Another emerging issue involves “prompt injections,” where hidden instructions embedded in digital content manipulate AI agent behavior. These techniques could be used to encourage self-preferencing, exclude competitors, or even coordinate pricing behavior between rival firms through covert signals. The authors also discuss risks linked to AI ecosystem “lock-in,” where dominant digital platforms may use proprietary AI agents to favor their own services and reinforce market power, raising concerns under abuse of dominance and digital market rules.

The article concludes that companies cannot avoid antitrust liability simply because an algorithm made the decision. Regulators are expected to focus heavily on governance, oversight, system design, and whether businesses could reasonably predict or prevent anti-competitive outcomes. To reduce risk, the authors recommend embedding antitrust compliance into AI systems from the outset, maintaining strong human oversight, carefully monitoring training data and prompts, conducting audits and risk assessments, and ensuring staff are trained to identify emerging competition-law issues tied to agentic AI.

CONTINUE READING…