AI and Consumer Markets: UK Competition and Consumer Law Implications for Businesses

The UK’s primary competition enforcement agency, the Competition and Markets Authority (CMA), recently published a discussion paper examining the competitive implications of agentic AI — autonomous systems capable of planning, deciding, and acting on behalf of consumers across multiple services — marking a significant escalation in the regulator’s scrutiny of AI-driven market conduct.

For those unfamiliar with the CMA’s work in this area, it is worth noting that the CMA explicitly identifies agentic AI as a potential step-change in how consumers interact with markets, with direct consequences for competition, consumer welfare, and the structure of digital ecosystems.[1] 

In-house counsel at businesses developing, deploying, or procuring agentic AI systems may wish to treat this paper as an early indicator of the CMA’s enforcement and market investigation priorities in this space.

1. What the CMA Is Saying: Key Themes and Proposals

• Agentic AI redefines the competitive interface. Agentic AI is different from prior generative AI tools: agents do not merely respond to queries but autonomously set goals, decompose tasks, retrieve real-time data (including personal data), coordinate with other agents, and take end-to-end actions — potentially across multiple services simultaneously. This fundamentally alters how competition for consumer attention and purchasing decisions operates.
• Concentration and gatekeeping risks at foundation model and platform layer. There is a concern that a small number of incumbents controlling foundation models, cloud infrastructure, or operating systems might leverage those positions to preference their own agentic services or foreclose rival agents — echoing the CMA’s existing strategic digital markets work and signalling potential overlap with the Digital Markets, Competition and Consumers Act 2025 (DMCC Act) regime.
• Consumer trust, manipulation, and information asymmetry. There is a concern that agents acting on behalf of consumers might be designed — or commercially incentivised — to steer users toward outcomes that serve the deploying business rather than the consumer. This includes personalised pricing exploitation, biased recommendation, and the risk that consumers lose meaningful oversight of consequential decisions taken on their behalf.
• Multi-agent coordination raises horizontal concerns. Where multiple AI agents interact — whether across a supply chain, in negotiations, or in pricing contexts — the CMA expressly raises the risk of algorithmically facilitated collusion or information exchange, even absent explicit agreement between the humans behind those systems. This signals a clear direction of travel toward scrutiny of inter-agent conduct under UK competition law.
• The CMA signals proactive, cross-regime engagement. The paper reflects coordinated engagement between the Financial Conduct Authority (FCA), the Information Commissioner’s Office (ICO), Office of Communications (Ofcom), and the CMA. The agencies are assessing both its existing powers (CA98, market investigations) and the DMCC Act’s new Strategic Market Status (SMS) and conduct requirement tools as potentially applicable to agentic AI ecosystems.



2. Competition Law Risks: Implications for Businesses

Risk 1: Algorithmic Collusion via Multi-Agent Systems

• Conduct under scrutiny: AI agents autonomously exchanging pricing signals, coordinating bids, or aligning commercial behaviour with competitor agents — without human instruction at the point of interaction.
• Legal exposure: under the Competition Act 1998 (the prohibition against restrictive agreements), The CMA’s established position is that algorithmic systems operated by businesses remain subject to the cartel prohibition; hub-and-spoke infringement theories and concerted practice analysis are both engaged where agents share or respond to competitively sensitive information.
• Severity/Urgency: High — the CMA has active prior enforcement interest in pricing algorithms, and multi-agent environments substantially increase detection difficulty and unintended coordination risk.



Risk 2: Abuse of Dominance via Agent Preference and Foreclosure

• Conduct under scrutiny: Dominant platforms or foundation model providers designing agent ecosystems that favour their own downstream services, restrict interoperability for rival agents, or apply discriminatory access conditions to APIs and data.
• Legal exposure: under the Competition Act 1998 (the prohibition against the abuse of dominance) and potentially DMCC Act SMS conduct requirements. Relevant theories of harm include self-preferencing, margin squeeze, refusal to supply, and tying/bundling.
• Severity/Urgency: High — the CMA’s existing digital markets precedents (e.g. mobile ecosystems and cloud markets) map directly onto these concerns, and the DMCC Act creates new, faster enforcement pathways.



Risk 3: Consumer Steering and Exploitative Commercial Design

• Conduct under scrutiny: Agentic systems designed to exploit consumer inattention, steer purchasing decisions toward higher-margin or commercially affiliated outcomes, or apply personalised pricing in ways that extract consumer surplus from identified vulnerable groups.
• Legal exposure: under the Competition Act 1998 (the prohibition against the abuse of dominance e.g. exploitative abuse); and under the UK’s unfair trading regulations, where the CMA’s consumer enforcement powers under the DMCC Act may be engaged.
• Severity/Urgency: Medium-to-High — the CMA has signalled this as a priority consumer harm; businesses combining dominance with opaque agent design may face compounded exposure.



Risk 4: Market Structural Concerns Triggering Investigation Powers

• Conduct under scrutiny: Accumulation of data, user relationships, and AI capability by a small number of firms creating structural features — barriers to entry, network effects, or data advantages — that may harm competition across multiple markets in which agents operate.
• Legal exposure: Enterprise Act 2002 market investigation regime; the CMA retains a broad discretion to investigate market features adverse to competition, and the paper signals this may be deployed if voluntary engagement with the industry proves insufficient.
• Severity/Urgency: Emerging — no immediate enforcement, but the paper’s analytical framing is consistent with pre-investigation scoping work.



3. Considerations for Businesses

Against Risk 1:

• Priority: Consider auditing any multi-agent deployment for data flows between agents — particularly pricing, inventory, or capacity data — and assessing whether those flows may constitute information exchange with a competitor’s system, even indirectly.
• Medium-term: Consider establishing AI governance protocols requiring legal sign-off before agents are permitted to interact with counterpart systems operated by competitors or commercially linked third parties.



Against Risk 2:

• Priority: Businesses that operate platforms through which third-party agents are deployed may wish to review API access terms and default ranking or recommendation logic for self-preferencing indicators that might attract Competition Act 1998 or DMCC Act scrutiny.
• Medium-term: Businesses may wish to engage proactively with a DMCC Act SMS designation risk assessment; those approaching the relevant thresholds should consider preparing compliance frameworks now.



Against Risk 3:

• Priority: Businesses may wish to review the commercial incentive structures embedded in any consumer-facing agent product — specifically whether agent reward functions or supplier relationships create misalignment between agent conduct and consumer interest.
• Medium-term: Companies should consider documenting the decision-making logic of deployed agents sufficiently to demonstrate, if required, that design choices were not exploitative — this may be essential in any regulatory engagement.



Against Risk 4:

• Medium-term: Businesses may wish to conduct a data asset mapping exercise to assess whether their AI and agent-related data holdings, combined with their market positions, might constitute a “feature” adverse to competition for market investigation purposes.



4. Next Steps and the CMA Process

The CMA is informally consulting with stakeholders and there is ongoing engagement with the FCA, ICO, and Ofcom, which indicates that findings will inform both its AI Foundation Models work and potential use of its Competition Act 1998 and DMCC Act powers. In-house counsel should monitor the CMA’s AI and digital markets publications page for response deadlines and follow-on announcements. No merger control filing thresholds specific to agentic AI are referenced in the paper.

---

[1] The UK’s Financial Conduct Authority has also published its own review on the implications for regulation and competition of AI and agentic AI in the UK financial services sector.