Data residency has two parts that vendors blur: where your data is stored at rest, and where the model actually runs on it (inference). As of September 2026, OpenAI offers data residency, including the United Kingdom and Europe, for ChatGPT Enterprise, Edu and the API, with in-region inference options for some Enterprise tiers, but not for ChatGPT Business. Anthropic’s own API and Claude plans store and run inference in the United States, with a global or US inference choice; Claude in EU regions is available through AWS Bedrock and Google Vertex AI under those clouds’ agreements. Gemini for Workspace follows Workspace’s data regions settings. UK GDPR does not require data to stay in the UK; it requires a lawful transfer mechanism, and the UK’s adequacy decisions and the standard clauses cover the main providers. Most UK SMEs therefore do not need in-region processing; regulated firms, public-sector suppliers and those whose clients contractually require it do, and for them the hyperscaler routes are how it is done.
- Two questions
- Where stored, where inference runs
- UK GDPR requires
- A lawful transfer mechanism, not UK storage
- Who needs in-region
- Regulated, public-sector, contract-bound firms
Storage and inference are different questions
A provider can store your conversations and files in a UK or European region and still send each request to servers elsewhere to run the model. Vendors’ residency pages often describe the first and are quiet about the second. When a client or a regulator asks where your data is processed, the answer has to cover both: where it rests, and where the model runs. Ask each provider both questions in writing and record the answers with the DPA.
The main providers, as of September 2026
Read from each provider’s own trust and residency pages. The distinction that matters is the third column.
| Provider and tier | Storage | Inference | Source |
|---|---|---|---|
| ChatGPT Enterprise, Edu, API | Residency available in UK, Europe, US and other regions | In-region inference options for eligible Enterprise tiers; otherwise US | OpenAI |
| ChatGPT Business | Not eligible for residency | US | OpenAI |
| Claude API, Team, Enterprise (Anthropic direct) | United States | Global or US selectable; no EU or UK option | Anthropic |
| Claude via AWS Bedrock | AWS region of your choice, including London and Frankfurt | Same region, under the AWS DPA | AWS |
| Claude via Google Vertex AI | Google Cloud region of your choice, including EU regions | Same region, under the Google Cloud DPA | Google Cloud |
| Gemini for Workspace | Follows Workspace data regions settings | Per Workspace terms |
What UK GDPR actually requires
UK GDPR restricts transfers of personal data outside the UK unless a mechanism applies: an adequacy regulation for the destination, the UK’s international data transfer agreement or the addendum to the EU standard clauses, or another safeguard. The UK has adequacy for the EEA, and the UK extension to the EU–US Data Privacy Framework covers certified US organisations. So a UK firm using a US-based AI provider is not breaking the law by doing so; it is making a restricted transfer that needs the mechanism recorded, a transfer risk assessment where the ICO expects one, and the provider’s position in the processor record. Storage location is a control, not a legal requirement in itself.
RelatedIs ChatGPT GDPR compliant? A UK guide Is Claude GDPR compliant? A UK guide
Who genuinely needs in-region processing
Four groups. Firms regulated in sectors whose regulator or client contracts require UK or EU processing, which is common in financial services, some health settings and legal work for certain clients. Suppliers to the public sector, whose frameworks often specify it. Firms whose own clients have written it into their contracts, which is increasingly common in questionnaires. And firms that have decided as policy that special-category data will not leave the UK or EEA, which is a legitimate choice. Everyone else has a transfer mechanism to record and a decision to document, and the choice of provider can be made on quality and cost.
How to get it when you need it
For ChatGPT, an Enterprise agreement with residency selected, or the API with the residency option, and confirmation in writing of whether inference is in-region for your tier. For Claude, the AWS Bedrock or Google Vertex AI route in a London or EU region, which is how we deploy for clients who need it, under the cloud provider’s DPA, with the model still Anthropic’s. For a built system, the choice is made in the architecture: the model endpoint, the storage and the logs all in the chosen region, stated in the runbook. It costs little more than the direct route and it answers the questionnaire in one line.
RelatedWhich Claude model for which job Your data when a supplier builds your software Custom AI solutions
Common questions
Does ChatGPT store data in the UK?
It can, for ChatGPT Enterprise, Edu and the API, where residency in the UK and other regions is offered, with in-region inference options for eligible Enterprise tiers. ChatGPT Business is not eligible.
Does Claude offer EU or UK data residency?
Not through Anthropic directly: its API and plans store and run inference in the United States, with a global or US inference choice. Claude in London or EU regions is available through AWS Bedrock and Google Vertex AI under those providers’ agreements.
Is it legal for a UK business to use a US AI provider?
Yes, as a restricted transfer with a lawful mechanism: adequacy where it applies, including the UK extension to the EU–US Data Privacy Framework for certified providers, or the UK transfer agreement or addendum, recorded with the DPA.
Do I need UK data residency for AI?
Only if a regulator, a framework, a client contract or your own policy requires it. Most UK SMEs need a recorded transfer mechanism, not in-region processing.
What is the difference between data storage location and inference location?
Storage is where your data rests; inference is where the model runs on it. A provider can offer UK storage while running the model in the US. Ask both questions in writing.
How do I get Claude with UK or EU processing?
Through AWS Bedrock in the London or Frankfurt region, or Google Vertex AI in an EU region, with the cloud provider as processor. Built systems can be architected that way from the start.