Independent: no vendor pays for coverage · Built from the DRCF statement and GOV.UK · Prepared with AI assistance; documentation review · How we test · Edited by Annie Vane
The business that deploys it stays responsible, according to the UK regulators. On 30 October 2026 the UK's four digital regulators (the CMA, FCA, ICO and Ofcom) said in a joint statement that existing law already applies to personal AI agents, and that "an AI agent's autonomy does not diminish the responsibility of organisations that deploy it". There is no new AI agent law and no new certificate. Consumer, competition, data protection, online safety, financial services, payments and cyber security rules apply as they already do. The regulators say they "will not hesitate" to use their powers where they reasonably suspect a breach that risks harm. If your firm puts an agent in front of customers, or lets one act on your behalf, don't expect "the AI did it" to work as a defence: the regulators say an agent's autonomy doesn't reduce your responsibility.
Update, 16 November 2026: the same four regulators have since written to ministers, saying firms launching personal AI agents should meet their obligations and engage with regulators before launch. See what the regulators' joint letter means.
Last checked: 8 November 2026 · Written by Annie Vane, editor · Evidence: Documentation review
Prepared with AI assistance; documentation review. No hands-on test. How we test
This page is about the four regulators' remits, the deploying firm's accountability and their six questions; our UK GDPR guide separately explains controller and processor responsibilities when an agent reads inboxes, calendars or customer data.
What the regulators actually said
The statement comes from the Digital Regulation Cooperation Forum (DRCF), through which the CMA, FCA, ICO and Ofcom co-ordinate. It is short (three pages): a position, not step-by-step guidance. The key points, in the regulators' order:
- They are preparing for personal AI agents. These are agents that act for a user, "such as making purchases and communicating with businesses". The regulators say they welcome the technology.
- Agents may "amplify and accelerate risks and introduce new ones". The statement names three examples: fraud and scams, personal information sharing not in line with user expectations, and misleading recommendations, decisions or information.
- Existing law applies. The regulators "want to be explicit" that competition, consumer and data protection law, online safety, financial services, payments and cyber security requirements all apply to personal AI agents, as do areas outside their remits such as product safety.
- Autonomy doesn't reduce responsibility. The deploying organisation stays responsible for complying with the law.
- They will act. Where they identify conduct they "reasonably suspect breaches the law and risks harm to consumers or businesses", they will use their powers.
- They want early contact. Firms are "strongly" encouraged to engage early and test products with the regulators, including through regulatory sandboxes.
The CMA's Agentic AI and consumers page on GOV.UK, first published on 9 March 2026, was updated on 30 October 2026 to link the statement. It sits beside the CMA's own guidance on how businesses can use agentic AI while meeting their legal obligations.
Who counts as "the organisation that deploys it"?
The statement doesn't define "deploy", so read it in plain terms. You are likely to be in scope if you:
- put an AI agent on your website or app that talks to, recommends to or buys for your customers;
- give an agent access to your systems so it can send emails, change records, book, refund or pay in your name;
- build a product that other people use as their personal agent.
The statement says firms "developing or deploying" personal agents should think about the questions below, so a developer selling the agent and the firm switching it on can both be in scope. How responsibility is split between them in a particular case is a question for the specific law involved and your contract. That's why the contract matters (see the supplier checklist).
If you're a consumer using someone else's agent, the statement is mostly about protecting you.
The six questions regulators want you to think about

The statement says firms "should think carefully" about how people can understand six things. Each one maps to a check you can do now.
| The regulators' point | The question to ask about your agent | Where to start |
|---|---|---|
| When and how the agent acts on someone's behalf | Can a customer tell when they're dealing with an agent, and what it's about to do? | Label the agent; show a summary before it acts |
| How authority to act is given | What exactly has the user (or you) allowed it to do, and can that be narrowed or withdrawn? | Per-task permissions; approval before paying, sending or deleting |
| How personal data is used | Does the agent's data use match what people would expect, and what your privacy notice says? | Your DPIA and privacy notice; the supplier's DPA |
| How commercial interests are communicated | If the agent recommends products, does it favour anyone who pays, and do users know? | Disclose paid placements or tie-ins in plain words |
| Who is accountable for its decisions | Is there a named person in your business who owns what the agent does? | Name an owner; keep an action log |
| How to get redress | When it goes wrong, how does a customer complain, get it fixed or get money back? | A route to a human, with a time limit you stick to |
None of these is a new legal test, but they show what the regulators want firms to think through.
What this means for a small UK business
1. Treat the agent's actions as your own. If your agent misleads a customer about price, sends marketing it shouldn't, or shares data in a way people didn't expect, assume it is treated as your firm doing it. For AI phone and message campaigns, our AI cold-calling guide covers PECR consent.
2. Get the supplier's answers in writing before go-live. The six questions above overlap heavily with our ten supplier questions: what the agent can access, whether there's an action log, how you switch it off and who processes the data.
3. Limit what the agent can do. Least privilege is one of the cheapest protections against both scams and mistakes. Read our AI agent security notes on prompt injection and third-party skills before connecting an agent to email, payments or files.
4. Keep a record. If a regulator or customer asks what happened, you'll need a log of what the agent did, when, and on whose authority.
5. Consider talking to the regulator early. The statement invites it. If you're building an agent product (not just using one), each member regulator's innovation services and sandboxes are the route they point to. The statement does not name which service fits a small firm, so check the relevant regulator's site.
What the statement does not do
- It creates no new law, duty or certificate. It restates how current law applies.
- It names no lead regulator for agents. Which regulator acts depends on the harm: data (ICO), consumer or competition (CMA), financial services and payments (FCA), online safety and communications (Ofcom).
- It sets no deadlines and no technical standards.
- It doesn't settle how liability is shared between an agent's developer, the firm deploying it and the user in a particular case.
The ICO's separate call for evidence on agentic AI is open until the end of 20 November 2026 (checked on the ICO's call page on 8 November 2026). Our ICO AI guidance tracker follows what the ICO publishes next.
Frequently asked questions
If an AI agent makes a mistake, is the business liable?
The regulators' position is that an agent's autonomy "does not diminish the responsibility" of the organisation that deploys it to comply with the law. So in practice, plan as if the agent's actions are your firm's actions. Whether you're liable in a particular case depends on the specific law and facts.
Is there a UK law specifically for AI agents?
Not as of 8 November 2026. The DRCF statement says existing frameworks apply: consumer, competition, data protection, online safety, financial services, payments, cyber security and others such as product safety.
Which regulator deals with a complaint about an AI agent?
It depends on the harm, and on which regulator has powers over it. For example, the ICO covers data misuse; the CMA covers misleading selling or unfair terms; the FCA covers payments and financial products; Ofcom covers online safety and communications. The statement says the four will share intelligence and co-ordinate.
Does this apply to agents I only use internally?
The statement focuses on personal agents acting for consumers and businesses. Internal agents still fall under existing law (data protection especially), and the same six questions are a sensible test.
Sources
- DRCF, Protecting people and businesses in an agentic economy: a joint statement by DRCF members, 30 October 2026 (landing page)
- DRCF, joint statement PDF, 30 October 2026 (quotes on this page are from this document)
- CMA, Agentic AI and consumers, GOV.UK, first published 9 March 2026, updated 30 October 2026
- DRCF, joint letter on personal AI agents to ministers, 16 November 2026 (checked 17 November 2026)
- Secondary, for context only: Burges Salmon, "AI agents and responsibility - a message from regulators" (burges-salmon.com)
This page explains UK regulator statements in plain English, checked against drcf.org.uk and gov.uk on 8 November 2026. It is general information, not legal advice. If an AI agent's actions could expose your business to a complaint or enforcement, ask a solicitor or data protection professional. Spotted something out of date? Tell me.



