Your notes and AI: privacy first

News
1

AI can be useful for summarising, rewriting and asking questions about your notes. But your notes are private, so you should always know where your data is going.

Joplin does not send your notes to an AI service by default. AI features are disabled by default, and you have to explicitly enable them and choose an AI provider. Learn more about AI Chat.

Keep your notes on your own device

You can connect Joplin to a local AI model, such as Ollama or LM Studio. In this case, your notes stay within your own device or private network and are not sent to an external service.

Joplin also blocks remote AI providers by default, as an additional safeguard.

You control what AI can access

Joplin lets you choose which tools are available to AI. This applies both to AI Chat and to the Model Context Protocol (MCP) integration.

For example, you can allow an AI assistant to search your notes without giving it permission to modify them. You can enable or disable individual tools depending on what you are comfortable with.

This means that connecting an AI service to Joplin does not automatically give it unrestricted access to your data or to all of Joplin's capabilities. Learn more about MCP.

If you use a cloud AI provider

You can connect Joplin to cloud AI services such as Joplin Cloud AI, OpenAI or Anthropic. When you do, the relevant data is sent to the provider you have selected.

For AI Chat, Joplin only sends the note content required for your request — such as the currently open note or selected text — rather than automatically sending your entire notebook. Learn more about the AI Chat panel.

You are in control

Joplin's approach is simple: AI should be opt-in, transparent and configurable.

You choose whether to use AI, which provider to use, whether to use a local or remote model, and which tools that AI is allowed to access.

If you want maximum privacy, use a local AI model and only enable the tools you actually need.

3

As it should be. Thanks for committing to security first. I much prefer a local LLM for my data. If any of you are not aware of what's being used, sold, and marketed when you use ANY AI cloud provider it's high time you spend some time researching it. Remember.. to them.. YOU are the product.

4

I'm a Joplin Cloud subscriber using Joplin Desktop 3.7 with Joplin Cloud AI (beta) as the chat provider, and I'd like to understand the data handling before using it with more sensitive notes. I couldn't find these points in the AI chat help page, the Plans page, or the Joplin Cloud privacy policy:

1. Model and vendor. Is Joplin Cloud AI a model you run on your own servers, or does it call a third-party provider? If third-party, which one, and where is it processed?

2. Retention and logging. Are prompts, note content sent as context, and model responses stored or logged? If so, for how long, and can I request deletion?

3. Training. Is any of this data used to train or fine-tune models, either by Joplin or by an upstream provider? Is there a zero-retention or no-training agreement with the provider?

4. Privacy policy scope. The policy says personal data is not shared with any third party. How does that apply to Cloud AI requests if an external model provider is involved?

5. Encryption. The chat panel refuses encrypted notes, so I assume Cloud
AI always receives plaintext. Is that plaintext decrypted only in
memory for the request, or stored anywhere server-side?

6. Plugins and tools. When a plugin such as Jarvis routes chat through
Joplin's built-in provider and includes excerpts from other notes, are
those handled under the same Cloud AI terms? And does the chat panel's
search/read-notes tool permission let it retrieve other notes, given
the help page says the panel never reads other notes?

Thank you.