The Nextcloud AI Assistant is a generative AI interface built directly into Nextcloud Hub, the open-source collaboration platform. It allows users to generate, summarise, translate and rewrite text, analyse documents and automate repetitive tasks, all from within the familiar Nextcloud interface. Because the Assistant connects to any OpenAI-compatible API endpoint, organisations can point it at a locally hosted model instead of a commercial cloud service, keeping every prompt and every response entirely within their own environment.
What the Nextcloud AI Assistant can do
The Assistant covers a broad set of daily productivity tasks. Its capabilities are not limited to a single context: they appear inside Files, Talk, Mail and other Nextcloud apps through a unified sidebar.
Practically, users can ask the Assistant to:
- Draft or rewrite documents and emails based on short instructions
- Summarise long files or meeting transcripts into bullet points
- Translate content between languages without leaving the document editor
- Answer questions about the content of files stored in Nextcloud
- Generate code snippets or convert natural-language descriptions into structured data
The feature set is extended through the Nextcloud Assistant app, available in the Nextcloud App Store, which adds a persistent chat interface and task history. Administrators can expose specific AI tasks to end users while restricting others, which is relevant for organisations that want AI assistance for low-sensitivity work but not for confidential projects.
Where the AI runs and where your data goes
The answer depends entirely on which backend is configured. Nextcloud itself does not provide inference capacity; it delegates requests to whichever API endpoint the administrator specifies.
| Backend configuration | Where inference runs | Data leaves your environment? | GDPR Article 28 processor agreement needed? |
|---|---|---|---|
| OpenAI API or similar commercial service | Provider’s cloud (typically US jurisdiction) | Yes | Yes, and US CLOUD Act exposure applies |
| Self-hosted open-weight model (Llama, Mistral) | Your own server or private GPU cluster | No | Only if the server is operated by a third party |
| Managed private AI (Swiss or EU data centre) | Provider’s dedicated GPU infrastructure | Only to the contracted processor | Yes, but within GDPR-adequate jurisdiction |
The European Data Protection Board has stated explicitly: “Organisations subject to GDPR Article 28 must ensure that any processor, including an AI inference provider, offers sufficient guarantees on technical and organisational measures.” (EDPB Guidelines 07/2020 on the concepts of controller and processor.)
This means that connecting Nextcloud Assistant to any external API, even a European one, triggers processor obligations. Connecting it to a model running on your own hardware does not, because no third party processes the data.
GDPR fines issued by EU supervisory authorities since May 2018 now exceed EUR 4.5 billion in total (GDPR Enforcement Tracker, CMS Law, 2024), which illustrates the financial stakes of getting data flows wrong.
How private GPU infrastructure makes sovereign AI practical
Running a capable large language model locally is not a configuration task; it is an infrastructure project. A model such as Mistral 7B or one of the Llama 3 variants requires several gigabytes of GPU VRAM just to load, and parallel user requests multiply that requirement quickly. Organisations without their own GPU capacity typically face a choice between slow CPU inference and sending data to a commercial provider.
A dedicated GPU cluster changes that equation. When inference runs on purpose-built GPU hardware, response latency drops to levels comparable to commercial APIs, making the assistant genuinely usable in daily workflows rather than a slow experiment.
Frank Karlitschek, founder and CEO of Nextcloud GmbH, has put it directly: “AI that leaks your data to a third-party cloud is not a productivity tool; it is a liability. Sovereign AI means the model, the inference and the output all stay within your legal and technical perimeter.”
Qsentinel runs exactly this architecture: dedicated GPU cluster infrastructure paired with managed Nextcloud Enterprise, with model serving handled for open-weight models including Mistral and Llama. The Nextcloud Assistant is pre-configured to point at that private backend, so no prompt ever routes through a commercial inference API. The hosting is located in Switzerland or on-premise at the customer’s own facility, both outside the scope of the US CLOUD Act.
Nextcloud now reports more than 400,000 server deployments globally (Nextcloud GmbH, 2023), and the share of European organisations expressing concern about data sovereignty when using US cloud AI services has reached 69% (European Commission Eurobarometer data, 2023). These two figures together explain why managed private AI is moving from a niche requirement to a standard procurement criterion for regulated sectors.
FAQ: Nextcloud AI Assistant and private AI
Does the Nextcloud AI Assistant send data to OpenAI or other external APIs?
Not by default. When configured with a local backend such as a self-hosted Llama or Mistral model, all inference happens on your own infrastructure. If you choose to connect an external API, that is an explicit configuration choice, and data does leave your environment.
Which AI models does Nextcloud AI Assistant support?
Nextcloud Assistant is model-agnostic. It connects to any OpenAI-compatible API endpoint, which means open-weight models such as Mistral and Llama can be served locally and used without modification.
Is a GPU required to run the Nextcloud AI Assistant locally?
Technically no, but CPU-only inference is slow for large models. Practical sovereign AI deployments use a dedicated GPU cluster to deliver response times comparable to commercial services.
How does private AI hosting relate to GDPR compliance?
Under GDPR Article 28, your AI inference provider qualifies as a data processor and must sign a Data Processing Agreement with appropriate safeguards. Hosting inference on your own infrastructure, or with a provider in a jurisdiction not subject to the US CLOUD Act, removes the cross-border transfer risk entirely.
What is the difference between Nextcloud Assistant and a managed private AI workspace?
Nextcloud Assistant is the built-in AI feature inside Nextcloud. A managed private AI workspace adds the GPU infrastructure, model management, security hardening and SLA on top, so the Assistant is fully operational without requiring the customer to build or maintain that backend.
Hoe Qsentinel dit oplost
Qsentinel is the managed Nextcloud Enterprise workspace, enhanced by Qsentinel with post-quantum encryption and sovereign private AI, hosted in Switzerland or on-premise, out of reach of the CLOUD Act.