Any AI agent that handles your personal data in the UK is covered by UK GDPR. The Information Commissioner's Office (ICO) says agentic AI processing remains "fully subject to UK GDPR obligations". In practice, you can find out what an agent provider holds about you, have it corrected or deleted, object to some uses, and challenge significant decisions made without a human. If the company does not help, you can complain to the ICO.

Why do AI agents raise data protection questions?

A chatbot answers questions. A personal agent acts. To book a table, rearrange your diary or reply to emails, it needs ongoing access to your inbox, calendar, contacts, files and sometimes your bank or health apps. It also builds a long-term "memory" of your preferences.

That creates three kinds of risk the ICO has flagged in its January 2026 Tech Futures report on agentic AI, which the ICO says is not formal guidance:

  • Using more data than a task needs. An agent booking a dentist does not need your whole inbox.
  • Special category data. Health, religion, sexuality and similar data can turn up in emails and calendars, and agents may disclose it.
  • Hard-to-explain processing. When an agent chains many steps together, it can be difficult to show what data was used and why.

The ICO's suggested mitigations include human approval before data access, data masking and logging what the agent did.

Each arrow in that chain is a point where personal data moves, and where UK GDPR applies.

Who is responsible for my data when an agent uses it?

Usually the agent provider (OpenAI, Meta, Anthropic, Amazon, Apple and so on) is the controller for the data it collects to run its service. That means it must have a lawful basis, tell you what it does with your data, keep it secure and respond to your rights requests.

When your agent passes your details to a shop or restaurant, that business becomes a controller for the data it receives. And if the agent emails other people, their data is involved too.

The ICO's call for evidence asks the industry exactly these questions about controller roles, so expect clearer guidance in 2027. See our explainer on the ICO call for evidence.

What rights do I have over an AI agent's data about me?

Right What it lets you do with an agent Notes
Right to be informed Know what data the agent collects and why Check the provider's privacy notice
Right of access Get a copy of your data, including stored memory Normally one month to respond
Right to rectification Correct wrong information the agent holds Useful if its "memory" is wrong
Right to erasure Ask for data to be deleted Not absolute; reasons must be given
Right to restrict processing Pause use of your data while a dispute is resolved
Right to data portability Move data you provided to another service Partial answer to memory lock-in
Right to object Object to processing based on legitimate interests, and always to direct marketing
Rights on automated decisions Safeguards for significant decisions made without a human Reformed from 5 Feb 2026

As of November 2026.

Getting a copy of your data

A subject access request (SAR) is the most useful tool. Ask the provider in writing, by email or through its privacy portal, for all personal data it holds about you, including conversation logs, connected-account data and agent memory. The ICO's right of access page explains the process.

Under the Data (Use and Access) Act 2025, the provider only has to make "reasonable and proportionate" searches, and can pause the clock if it genuinely needs you to clarify your request.

Deleting memory

Most agents let you view and delete memory in settings, which is quicker than a formal request. If you want everything gone, including backups and data shared with partners, make a written erasure request. Our sister site compares agent memory and privacy controls.

What changed with the Data (Use and Access) Act 2025?

The main data protection changes in the Data (Use and Access) Act 2025 took effect on 5 February 2026, under SI 2026/82. Three matter for agents:

Most DUAA data protection changes in force
5 Feb 2026
Complaints-to-controller duty in force
19 Jun 2026
Usual deadline for a subject access request
1 month
ICO agentic AI call for evidence closes
20 Nov 2026
  • Automated decisions. The old Article 22 rules were replaced. Organisations can now use automated decision-making on a wider range of lawful bases, but for significant decisions you are still entitled to safeguards: information about the decision, the chance to make representations, human intervention and a way to contest it. Stricter limits remain where special category data is used.
  • Recognised legitimate interests. Some purposes, such as fraud prevention and security, no longer need a balancing test, though the processing must still be necessary.
  • Complaints. From 19 June 2026, organisations must have a process for handling data protection complaints. In practice, you should complain to the company before going to the ICO.

How do I use my rights with an AI agent?

  1. Check the privacy noticeFind out what the agent collects, who it shares data with and where it is stored.
  2. Use in-app controls firstReview connected accounts, memory and history, and delete what you do not want kept.
  3. Make a written requestSend a subject access, correction or erasure request to the provider's privacy contact.
  4. Complain to the companyIf the reply is late or unsatisfactory, use its data protection complaints process.
  5. Escalate to the ICOIf you are still not satisfied, complain to the ICO at ico.org.uk.

Does UK GDPR explain why agents are not available in the UK?

Partly, perhaps. No vendor has officially blamed UK GDPR for keeping Meta Muse, OpenAI dots on Pro plans or Claude Duet out of the UK. Commentators point to UK GDPR compliance work and the ICO's open review as likely factors. On paper, the 2025 Act made the UK more permissive than the EU on automated decisions, which suggests vendor caution rather than a legal block. We cover this in why AI agents are US-only.

If you get round the restriction with a VPN, your data may be handled under US terms rather than the provider's UK and EU commitments, which can make exercising these rights harder.

Practical privacy tips for agent users

  • Connect the minimum. Give an agent your calendar before your whole inbox, and read-only access before write access.
  • Avoid "always allow". YouTuber Matt Robb said Meta Muse shared his address in a Facebook Marketplace sale, and later admitted he had granted the permission that made it possible (TechCrunch).
  • Keep health and finance separate unless you really need the agent to use them.
  • Review connected apps every few months and remove ones you no longer use.

Our AI agent safety guide has more.

Bottom line

UK GDPR already gives you real rights over what AI agents do with your data: access, correction, deletion, objection and safeguards on automated decisions. The ICO is now working out how those rules apply to agents in detail. Until its guidance arrives, the best protection is to connect as little as possible and to use your access and erasure rights if something looks wrong. The DRCF joint statement confirms the ICO is watching.