Adopting consumer facing AI agents: CMA guidance for UK businesses

The Competition & Markets Authority (CMA) has released guidance for businesses regarding adoption of Agentic AI in consumer facing contexts. Within the UK, AI is being increasingly adopted by businesses to streamline workflows and enhance customer experiences. Within its guidance, the CMA acknowledges the UK’s position as the third largest AI market globally but emphasises the need for clear governance, strategy and assurance of compliance within AI adoption. This guidance provides crucial reminders surrounding liability for AI agents, consumer law protections to consider and how to mitigate risks of AI adoption.

What is Agentic AI?

Agentic AI does not have a specific individual quality or technical definition. Instead, it is understood as operating differently to traditional AI models which act on pre-defined, automated processes. For example, a commonly adopted use of traditional AI is retail chat bots, which have a limited role responding to consumer questions within defined pre-set answers. Alternatively, Agentic AI acts with increased flexibility, with the ability to make judgments based on the surrounding context and focuses more on outcomes rather than operating a task in isolation. Examples include independent sourcing of deals for consumers, autonomously making payments on a consumer’s behalf, or assessing goals and summarising necessary tasks to achieve these.

​​​​​​How businesses are using Agentic AI in consumer facing services

Within the UK, there has been an increase in the use of Agentic AI but currently adoption remains largely outside of the consumer sector. However, with the potential efficiencies and benefits including hyper-personalisation of consumer experience, lower prices and bespoke deals, it is anticipated businesses will look to adopt this next. Alongside the potential benefits, the guidance emphasises the importance of safeguards to ensure consumer trust and confidence and minimise hallucinations.

A hallucination is when AI generates an inaccurate output (e.g. a legal case which does not exist). Hallucinations are considered a major threat when using AI, with consumer watchdogs such as Which? warning of hallucinations they have found in generative AI programmes. This issue is particularly pertinent in the consumer context, if hallucinations lead to consumers purchasing goods or services they otherwise wouldn’t have. Consumer law restrictions also apply to consumer facing AI agents. Consumer law provides key protections including restrictions on misleading consumers, omitting material information and transparency requirements. Therefore, platforms where AI agents are misleading or manipulating consumers, by providing incorrect information, undue pressure and/or using deceptive design methods (e.g. having a small cancellation box in the corner of the user’s screen), would be in breach of consumer law. Crucially, businesses are responsible for any AI agent’s actions.

Whilst the CMA has emphasised it does not agree with a one-size-fits-all regulation for AI use, it is focused on identifying benefits and risks of AI and monitoring where further protections would be beneficial. The guidance emphasises the importance of carefully using AI in consumer facing settings and reminds businesses to consider the purpose of AI systems purchased and how to mitigate potential risks.

Key considerations referenced for businesses when implementing AI include:

• Ensuring regular human oversight and test the AI system frequently to spot any errors or biases and unintended consequences;
• Training the AI system on accurate information and on consumer law. Systems are only as useful as the data they are built on and the CMA references data mobility as a potential benefit to Agentic AI, to avoid biases and provide the most accurate offers, etc., to consumers; and
• Clearly communicating to consumers where and how AI is used, including clear labelling of AI content.

Ensuring compliance with UK consumer law

The CMA warns that any use of AI needs to be compliant with consumer law and cites the CMA key principles including transparency and accountability. Crucially, fines for breaches of consumer law can be up to 10% of global turnover under the Digital Markets, Competition and Consumers Act (DMCCA). For any use of AI in the EU, please be aware of the additional restrictions under the EU AI Act.

Alongside this, it is not clear how certain laws will apply in the context of AI. For example, Which? has queried if the protections offered by Section 75 (the credit card company’s joint and several liability for breach of contract or misrepresentation by the trader) and chargeback protections would apply to purchases made on an AI platform or following an AI recommendation.

Next steps for businesses adopting consumer facing AI

If you need advice on the deployment of AI in consumer contexts or general consumer law advice, please get in touch with the Commercial and Technology team at [email protected].