Provider switches
In Settings, under Organizations and Provider API keys.
Both start off. While China is off, GATE never sends your data to a provider in China, not even as a fallback.
GATE's sovereign AI keeps your agent's chats, memory, files and backups on servers in Germany and Finland, run by GATE Intelligence (Gate Intelligence AB, 559592-1395) in Stockholm under a published DPA. Providers in China and providers that may train on your data stay off until your owner allows them. In the default setup, the AI models run in the US under the EU's Standard Contractual Clauses.
GATE's servers in Germany and Finland hold your data. In the default setup, AI model calls go to the US, and providers in China stay off until your owner allows them.
Where your data lives, which AI providers may see it, and who decides.
Your agent's chats, memory, files and the database behind them run on GATE's servers in Nuremberg, Germany. Backups are encrypted and kept in Falkenstein, Germany, and support servers run in Helsinki, Finland.
Two switches in Settings decide whether your data may reach providers in China and providers that may train on what they receive. Both start off until an organization owner turns them on.
Your agent sends an AI model only the context of the task in front of it. GATE does not use your content to train AI models, and the search index of your agent's memory is built on GATE's own servers.
Calls to AI models in the US rely on the EU's Standard Contractual Clauses, plus the EU-US Data Privacy Framework where the provider is certified. GATE's subprocessor page names every company, where it sits and when your data reaches it.
If every step must stay in the EU, GATE Intelligence scopes an EU-only setup with you. A company can also run on Stable or on a dedicated server, where GATE holds no standing access to its data.
Your organization owner, with two switches that start off.
In Settings, under Organizations and Provider API keys.
Both start off. While China is off, GATE never sends your data to a provider in China, not even as a fallback.
Built into how GATE runs.
Each part of GATE's default setup, where it is processed and on what basis.
Your agent's chats, memory, workspace files and database are stored on GATE's servers in Nuremberg, Germany.
Backups are encrypted before they leave the server, kept in Falkenstein, Germany, and roll off within about 90 days.
Everyday chat and code sessions run on AI models in the US, under the EU's Standard Contractual Clauses and the Data Privacy Framework where certified.
Transcription, meeting speech, images and video use US providers when you use them, and sign-in emails go through a US email service.
Never used, not even as a fallback, until an organization owner turns on Allow providers in China.
A provider GATE classifies as training on what it receives runs only after an owner allows it. The DPA names each one, plus two the switch does not cover yet.
Three ways a European company can put AI to work on its own data, side by side.
| What you compare | A US AI assistant | A self-hosted open model | GATE's sovereign AI |
|---|---|---|---|
| Where your data is stored | In the provider's cloud, in the regions its plan offers | On servers you run yourself | On GATE's servers in Germany and Finland |
| Who runs it | A US company | Your own team | GATE Intelligence, a Swedish company |
| Which AI models | The provider's own | The open models you install | Several, picked per task, with providers in China and training providers off by default |
| What leaves the EU | Depends on the provider and your plan | Nothing, if your servers are in the EU | AI model calls in the default setup, under SCCs; an EU-only setup is scoped on request |
| Who can read your data | The provider, under its own terms | Your own team | Named GATE Intelligence staff on Edge; no standing access on Stable or a dedicated server |
| Who keeps it running | The provider | Your engineers | GATE Intelligence |
| What it does for you | Answers in a chat window | What your engineers build around it | Your own agent, with tools for the books, contracts, posts and support |
| Data processing agreement | The provider's standard terms | None needed, nobody processes for you | A published DPA under Swedish law |
Most companies run one GATE agent, named whatever they like. Its chats, memory and files live in the EU, and your owner's switches decide which AI providers it may use.
Meet your agentSovereign AI is AI whose data, operator and rules stay under a jurisdiction and an owner you choose, instead of wherever a provider happens to run. GATE offers it to European companies: your agent's chats, memory and files are stored in Germany and Finland, GATE is run by Swedish GATE Intelligence under a published DPA, and providers in China or that may train stay off until your owner allows them.
Yes. GATE runs on servers in the EU: your agent's chats, memory, files and database sit in Nuremberg, Germany, encrypted backups in Falkenstein, Germany, and support servers in Helsinki, Finland. The AI models your agent thinks with are a separate hop: in GATE's default setup they run in the US, under the EU's Standard Contractual Clauses.
Some of it does in GATE's default setup. The context of each task goes to AI models in the US, and reply quality checks, transcription, images, video, sign-in emails, edge traffic and integration tokens use US providers, under the EU's Standard Contractual Clauses plus the Data Privacy Framework where the provider is certified. Storage stays in Germany and Finland, and GATE's Trust Center lists every flow and its legal basis.
Only with an EU-only setup agreed for your company. GATE Intelligence scopes a local EU package with you, with AI models, email, data and hosting on EU-based providers. Until it is agreed and set up, your organization runs on GATE's default setup, where AI model calls go to the US under the EU's Standard Contractual Clauses.
No. GATE does not use your content to train AI models. A provider GATE classifies as training on what it receives is used only after your organization owner turns on Allow providers that may train on your data, which is off by default. GATE's Data Processing Agreement names each such provider, and the two whose terms allow training by default where this switch does not yet apply.
Not unless your organization owner allows it. GATE's switch Allow providers in China is off by default, and while it is off GATE never sends your data to a provider in China, not even as a fallback when its main AI model is unavailable. An owner can turn it on in Settings, and GATE's subprocessor page names the providers it opens up.
It depends on where your organization runs. On Edge, GATE's default release track, named GATE Intelligence staff and the bots they run can open your chats, files and memory to operate, support and improve GATE. On Stable, and on a dedicated server of your own, GATE holds no standing access and opens your content only when you ask, case by case. Other organizations' agents cannot reach your content either way.
GATE Intelligence, a Swedish company, publishes a Data Processing Agreement under GDPR Article 28, governed by Swedish law. Transfers outside the EU rely on the EU's Standard Contractual Clauses, and a breach that meets the GDPR threshold is reported to IMY, the Swedish data protection authority, within 72 hours, with affected customers told without undue delay. GATE's Trust Center maps its controls to GDPR, ISO/IEC 42001 and the SOC 2 criteria.
GATE is invite only right now while GATE Intelligence onboards a small number of businesses. Request an invite on this page, and if your company needs a signed DPA, an EU-only setup or a dedicated server, email legal@gate.software so it is agreed before your data goes in. Your organization owner can set the provider switches from the first day.
Invite only while we onboard a small number of businesses.