Settings › Security
Who can do what, two-step, AI tools, data region, audit log, export and deletion
On this page
Settings › Security is for workspace admins. Other privacy controls are elsewhere: how long raw transcripts are kept, masking and clearing people's personal details are in Settings › Privacy (a product admin's; see Conversations and People), and your own two-step sign-in is in Settings › You › Sign-in.
Who can do what
A table of what Admins, Editors and Viewers can do, read from the app's own permission checks, so it is always what the app does. Workspace viewers and product viewers never take a seat. An AI tool's "Read and update" token can only change what its owner can, checked on every call.
Two-step sign-in
Who has it on lists, by name, who uses two-step and who doesn't. Require two-step sign-in for everyone is a switch; turn it on for yourself first (Settings › You › Sign-in). Once the new MindBacklog is on for everyone it is required at each person's next sign-in; until then each person still turns it on for themselves, and the page says so. Two-step means a six-digit code from an authenticator app, plus 8 recovery codes.
AI tools signed in to this workspace
Each tool someone signed in (Claude Code, Cursor, Lovable and others) with who, how it signed in, what it can do (Can read, Can read and update status, Can read and update), which product, when it was last used and since when. It acts as its person with their access, checked on every call. Sign out stops it on its next call. Personal tokens are listed in each person's Your AI tools.
Let AI tools update status (off by default) is for editors on products they can only read: their AI tools can tick off work items and mark features shipped, and nothing else. A viewer's AI tools only read. People who can edit always can.
Data region
Where your data is stored is not confirmed yet, so the page says Not confirmed yet. We'll state it here once we have checked it. There is no EU region today. I need the EU records why and by when (both optional) so MindBacklog knows who needs one; it changes nothing for you today. Mind's AI models are run by the providers listed under Who processes data for us, where they run them, whatever region stores your data.
What's true today
Where your data is stored, how connection secrets are kept (encrypted, never shown again after you connect), the nightly backups (02:30 UTC; ten daily and two weekly copies), what deletion leaves behind, exports, the audit log, single sign-on (planned; Google and Microsoft sign-in work today), SOC 2 (not yet) and Who processes data for us. Anything not confirmed says Not confirmed yet rather than making a claim.
The audit log
Every change, newest first, kept twelve months: who changed roles, members, teams, AI tool tokens, connections and the plan, and what AI tools changed. 50 lines a page, with a filter by what happened, and Download it as CSV. Times are UTC.
Export everything
A zip of the workspace's data: a CSV for each kind of record, the members, and every document as Markdown. Secrets and search indexes are never in it. The admin who asked gets an email with a link, usually within 10 minutes; the link works once, for that admin, and the file is deleted after seven days. The page shows the last export's state (waiting to start, being made, ready, failed, expired). One export an hour, and one runs at a time.
Delete the workspace
Type the workspace's name to confirm. Members lose access at once, and everything is deleted seven days later. Every admin gets an email with a Cancel link, and any admin can cancel during those seven days; everyone else sees a page saying the workspace is being deleted and when. Copies in the backups are removed as the backups age out (up to about three weeks).
FAQ
Can an editor or viewer open Security? No. It is for workspace admins.
Where do I export one person's data? On their page in Learn › People (a product admin).
Where are my customers' personal details cleared? Settings › Privacy › Removing data for good.