Albert

// Where your data goes

Your work stays yours.

Every company that receives data from Albert, what it receives, how long it keeps it, whether it trains AI on it, and who owns it. Hand this page to your client, your auditor or your POPIA file.

Last updated 28 September 2026, from each company's published terms on that date. Applies to the Albert app version 0.17.32 and later, with the settings in Settings → Privacy switched on (they are on by default).

Who owns your work

You do. None of the companies below claims ownership of what you send or what the AI gives back, and neither do we. What differs is whether they may keep it or train AI on it — that is what the table shows.

The one thing to check is your Claude plan. On Claude Team, Enterprise or an API key, Anthropic never trains on your work, treats it as your confidential information, and defends you against copyright claims over what Claude writes. On a personal plan (Free, Pro or Max) Anthropic may train on it unless you switch "Help improve Claude" off, and gives no such protection. Albert shows which plan you are on in Settings → Privacy → Your Claude account.

Who receives what

Who What they receive How long they keep it Trains AI on it?
Anthropic (Claude, Claude Code) — on your own Claude account Conversation with Albert, builds, plans, meeting notes, and meeting answers Albert has to check in the project. For builds, the files Claude Code opens. Business plans: 30 days. Personal plans: 30 days with training off, up to 5 years with it on. Anything flagged by their safety checks: up to 2 years. Business plans: never. Personal plans: only if "Help improve Claude" is on.
OpenRouter and the AI companies it routes to (Google Vertex, Microsoft Azure, Amazon Bedrock, TypeSafe and others) Albert's helpers: quick meeting answers and question cards, email write-ups, reading a shared screen, small decisions, and pictures in designs. Only the part the task needs, with personal details hidden. Nothing is stored: only providers that keep no data are used, and a request no such provider can serve is not sent. Their terms still allow a short-lived memory cache. OpenRouter keeps a record that a request happened, not its content. No. Providers that train are switched off on our account as well as in the app.
Google (Gemini) — only if you add your own Gemini key The live talking voice and what it sees: your voice, and your camera or screen when you share them. Set by your key. A paid key: up to 55 days, for abuse monitoring. Paid key: no. A free key: Google may use it to improve its products, and people may review it. Use a paid key for client work.
Microsoft (Edge speech) The words of Albert's spoken replies, turned into sound. Personal details are replaced first ("the email address"). Not published by Microsoft for this free service. Not published. Only Albert's own replies are sent, never your files.
Swiftnex Tech (our servers, hosted by Supabase) Your account and team; on a team plan, the plans, meeting notes, emails and designs you choose to share; the board; a usage record. Until you delete it or close your account. No. We do not train AI on your work.

Two features run on our servers: a client chatting about a plan you shared, and drafting board jobs from shared meetings. They send the plan or notes to AI through OpenRouter without hiding personal details. Providers that train are blocked on our account, but these two do not yet ask for zero data retention.

What never leaves your computer

The settings behind this page

Each company is responsible for keeping to its own terms; we have not audited them. The full explanation is on the privacy page. Questions about how your data is handled? Write to hello@swiftnextech.com.