Privacy & what is sent

Lino AI reads what it needs to answer you, and what it reads is sent to the AI model. This page explains what goes out, and the switches that keep things away from it.

What is sent to the model

When you send a message, the model receives:

your message and the chat so far;
what you attached or mentioned (chips, files, pictures, quotes);
a short description of where you are — the current Canvas's name and path, its Canvas memory, and your saved memories;
whatever Lino AI reads while working: Canvases, Artifacts, files in folders you gave it, web pages, your Schedule.

Model requests go through Lino's own server to the model provider; your sign-in, not a key stored in the app, authorizes them. The context ring in the AI panel shows what share of the context each part takes (see The AI panel).

Hide from AI

Mark a Card, Image, Link, Sketch, Table or Spark Hide from AI and Lino AI stops reading it.

Turn it on

Right-click the object (on the Canvas, in the Artifacts Gallery, or in the Inbox) and choose Hide from AI. With several objects selected, choose Hide all from AI.
Or open its Properties panel and tick Hide from AI — Lino AI will not read this, or the name other objects use for it. From the next message on.

Choose it again to turn it off. Only you can change this setting; Lino AI can't.

What it hides

The object itself — its content, Properties and pictures. Lino AI is told only that something is hidden there.
Its name elsewhere — when another Card mentions it, Lino AI sees a placeholder instead of the name.
Its Markdown file — for a Card synced to a file, that file is refused to Lino AI's file tools and commands.
Suggestions — you get no inline autocomplete while typing in it.

On the Canvas it keeps its place and wears a small mark in its top-right corner (Hidden from AI).

When a message would carry it

If something you are about to send includes a hidden object — a mention, a selection, a chip — Lino stops and shows This is hidden from AI, with the object's name. Choose Remove and send to send the message without it, or Back to change it. Hidden items in the chip preview are struck through.

Limits

It takes effect from the next message. Anything already in a chat stays there.
Words you type yourself are yours to send — Hide from AI doesn't scan your message.
It hides a file by its path, not every copy. If the file is in a git repository, git's copies (git show, git log -p) are still readable, and so are backups and other copies. To keep a file from Lino AI entirely, keep it out of git and out of any folder you give it — or use .linoignore.

.linoignore

To keep files inside a folder you gave Lino AI away from it, put a .linoignore file in that folder (or any folder above). It is written like a .gitignore:

diary.md
private/
*.key

Lino AI cannot read, list, search or change anything it names — not with its file tools and not with any command it runs. It can't read or edit the .linoignore itself either. If it runs into one, it is told excluded by .linoignore and leaves the file alone or asks you to change the list. For the Canvas's working folder and scratch folder, see Canvas Binding.

Settings that limit what it reaches

All in Settings, under Lino AI:

Setting	Where	Default	Off means
Web access	Capabilities	On	No web search, page reading or importing from links. (Commands that reach a site still ask you either way.)
Read calendar events	Capabilities (desktop app)	On	Lino AI can't see your calendar events — titles, times, places, invitees.
Include notes and invitees' addresses	Capabilities, under the above	Off	Event notes and other people's email addresses are left out.
AI captions for new Images	Capabilities (Pro and above)	On for Pro	New Images aren't sent to the model to be named and described.
Generate memory from chats	Memory	On	Nothing new is saved from chats; what's already saved is still used.
Site logins	Permissions → General	On (asks per site)	Lino AI never asks to borrow a site login.

AI captions

On Pro and above, AI captions for new Images gives every new Image a name and a description of what is in it, text included: When an Image comes in, Lino AI names it and writes down what is in it, text included. The file name is only a hint. Images are sent to the AI model, and this uses AI credits. It runs for pictures you drop, paste, import or save from Quick Entry.

To caption one again (or caption one on a plan without automatic captions), right-click an Image or Sketch and choose Caption with AI.

Separately, when Lino AI reads a picture for the first time it writes a short description into the Image's Description Property, so it (and search) can understand it in words later. You can edit that description; your text is never overwritten.

Site logins

Some commands need you to be signed in — subtitles only an account sees, a page behind a login. Lino AI can ask to use your login from Lino's own browser (the Browser tab in the Right Panel) for that one site:

The card says Use your login for a command? — commands in that chat can act as you on that site until the chat ends.
If you aren't signed in there yet, the card offers Open site; sign in, then press I've signed in.
Only that site's cookies are handed over, readable only by that chat's commands, and deleted when the reply ends. Your other browsers' logins are never used.

Turn this off entirely in Settings → Permissions → General → Site logins.

Content from outside is not trusted

Text that comes from web pages, files, command output or comments is treated as data, not instructions. The web tools refuse localhost, private network addresses and URLs assembled from pieces, so a page can't point Lino AI at your own network. The Auto mode review never sees such text (see Modes & permissions).