
Key takeaways
- The CMA says you're responsible for what an AI agent does in the same way you're responsible for an employee, and consumer law breaches can cost up to 10% of worldwide turnover.
- Its guidance asks businesses to tell customers when an AI agent is used, train it on consumer law, monitor it and fix problems quickly.
- The ICO's call for evidence on agentic AI closes on 20 November 2026 and will feed into future guidance and a statutory code of practice.
- The practical steps overlap: clear permissions, a named owner, logs you can replay and a human who can change the outcome.
What changed in late October
On 30 October 2026 the Competition and Markets Authority updated its guidance Using AI agents: complying with consumer law, first published in March. The update links to a joint statement from the four members of the Digital Regulation Cooperation Forum (the CMA, FCA, ICO and Ofcom) on protecting people and businesses as personal AI agents spread.
The message is that existing law already covers agents. If yours handles refunds, price questions or bookings, the rules for human staff apply to it too.
What the CMA expects from businesses using AI agents
It sets out five areas:
- Same rules for AI and human agents. The CMA says you are responsible for what an AI agent does “in the same way you are responsible for what an employee does”. That includes agents bought from a third party.
- Tell customers. Say when an AI agent is involved where it could affect their decisions, and don’t overstate the AI’s role.
- Train it on consumer law. Prices should include unavoidable charges, offers must be genuine and cancellation rights must be respected.
- Monitor it. Test before launch, keep a human in the loop and have someone with the right experience review complaints, feedback and what the agent tells the public.
- Fix problems fast. Refine prompts and workflows quickly, especially when the agent reaches many people or vulnerable customers.
The worked examples are worth reading. One is an agent giving wrong information about cancellation rights. Another handles refunds, where it has to respect the Consumer Rights Act 2015, the Consumer Contracts Regulations 2013 and any longer returns window you offer. Small firms often hand exactly these tasks to a support chatbot first.
What the ICO is asking about
The Information Commissioner’s Office opened a call for evidence on agentic AI in October. According to law firm TLT, it runs for six weeks and closes on 20 November 2026. Freshfields’ summary says responses will inform future ICO guidance and a statutory code of practice on AI and automated decision-making.
The ICO’s early expectations, as Freshfields reports them, include:
- Give each agent a clear identity, a named owner, a stated purpose and set permissions, with approval thresholds that match the risk.
- A single notice at the start of a workflow may not be enough transparency. Extra prompts may be needed.
- Whoever sets the agent’s permissions and access may carry controller responsibilities.
- Human review has to be meaningful, meaning the reviewer can actually change the outcome, and logs should let you reconstruct what the agent did.
- An agent finding a new use for data doesn’t make that use compatible with the original purpose.
A short checklist for UK businesses
| Area | What to put in place |
|---|---|
| Ownership | One named person accountable for each agent |
| Permissions | A written list of what the agent can read, change and send, and what needs approval |
| Disclosure | Clear wording that customers are dealing with an AI agent |
| Consumer law | Prices, offers, refunds and cancellation answers checked against your terms |
| Logging | Records detailed enough to replay a decision |
| Escalation | A fast route to a person, and a way to pause the agent |
In our view, if you switched an agent on with default settings, review it now. Our guide to chatbot handoff and disclosure covers several of these points in practice.
What it means if you also sell into the EU
UK rules sit alongside the EU AI Act rather than replacing it. If you serve customers in both markets, plan one set of controls that satisfies both. Our article on the EU AI Act Digital Omnibus explains the EU side, and our UK and Europe consultancy page shows how we approach projects there.
If you’re planning an agent, we build them with permissions, logging and handoff from day one. See our AI agent development service or book a strategy call.
This article is general information, not legal advice. Speak to a qualified UK lawyer about your situation.
Frequently asked questions
Am I responsible if my AI agent breaks UK consumer law?
Do I have to tell customers they're talking to an AI agent?
When does the ICO's agentic AI call for evidence close?
What's the first practical step?
Sources
- GOV.UK (CMA): Using AI agents: complying with consumer law (updated 30 Oct 2026)
- TLT: The ICO’s agentic AI call for evidence, what you need to know (12 Oct 2026)
- Freshfields: AI agents and UK data protection law (22 Oct 2026)