Two kinds of residency
OpenAI separates two things people both call “keeping data in the EU”.
| Aspect | Data residency | Inference residency |
|---|---|---|
| What | Where your data is stored at rest | Where the model processes your data on GPUs |
| Who | New ChatGPT Enterprise and Edu customers, eligible API customers | Enterprise and Edu with data residency in the same region |
| Regions | Australia, Canada, Europe (EEA + Switzerland), India, Japan, Singapore, South Korea, UAE, UK, US | Europe (EEA + Switzerland), US, UAE |
| Cost | Included with Enterprise and Edu | Not priced separately on the page |
Source: OpenAI’s data residency help article, read in November 2026. Enterprise bought through AWS Marketplace can use any storage region but only US inference residency.
What stays in the region, and what does not
With data residency, this content is stored in the chosen region, including backups and replicas:
- ChatGPT Memory (stored conversation context).
- Conversations: text, image and voice.
- Uploaded files.
- Custom GPTs with their prompts and outputs.
- Code Interpreter and data analysis artifacts.
- Image generation inputs and outputs.
What can live outside the region: data sent to external integrations such as apps, MCP servers or web search, workspace name, billing information, user logins and other metadata. Beta features, Codex Web and third-party GPTs are not covered. Workspaces with data residency cannot set GPT sharing to “Anyone”.
And ChatGPT Business?
The residency article names Enterprise, Edu and the API only. As of November 2026, OpenAI’s business pricing comparison lists data residency as unavailable for Business and available for Enterprise. What Business does have:
- No training on your data, by default.
- A data processing addendum to support GDPR.
- AES-256 encryption at rest, TLS 1.2+ in transit, and a SOC 2 Type 2 audit.
- Admin-controlled retention; deleted conversations are normally removed within 30 days, except where longer retention is legally required or reasonably necessary to protect the service or others from harm.
If your data must provably stay in Europe, OpenAI’s own documentation points to Enterprise. See ChatGPT Team vs Enterprise.
The common mistake: EU on, everything in the EU
Even with both data and inference residency, not everything stays in the region. OpenAI says CPU-based processing such as extracting text from PDFs, request routing, authentication and analytics may happen elsewhere. And once ChatGPT calls a connected app or a search provider, that provider’s terms apply. If all processing has to stay inside your walls, a region setting is not enough.
What it means for your team
Where your data lives is a question about more than ChatGPT. A team that also uses Claude, Copilot or its own agents has a different storage location and a different rulebook per tool. Before giving AI your company knowledge, know which tool keeps what. On training and retention specifically, see does ChatGPT Team train on your data.
Where Bombos fits
Bombos keeps your team’s shared memory in one place you choose: in the cloud, or standalone on your own server with your own models. Hosting the memory yourself does not change where a connected cloud AI service processes prompts; use your own models as well if processing must remain on your own server. Every AI tool on the team reads from it, and you can see what each answer rests on. One person can try it first for their own agents.
Related questions
Can an existing workspace move to the EU?
OpenAI's article talks about new Enterprise and Edu customers and workspaces provisioned with data residency. For an existing workspace, ask OpenAI sales.
Is ChatGPT Business GDPR compliant without EU storage?
OpenAI offers a DPA for Business to support GDPR compliance. Whether that is enough depends on what you process; EU storage is not a GDPR requirement in itself, but it is an extra assurance.
Where can I see my workspace's region?
As an owner or admin: profile menu, Workspace settings, General, then the Data Residency Region field.
