“Can the assistant send something without our approval?”
Your approval comes first
Your assistant has its own name and email address, and it never sends messages as you. Anything that should go out in your organization’s name, to donors, funders, partners or the public, is a draft you review and send yourself.
Your assistant researches, drafts, organizes and keeps track of the work. A person decides what goes out. For some work, sending can be switched off entirely, so your assistant delivers finished work to your portal instead.
- Drafts first
- Named approvers on your team
- Checks on every outgoing message
- Approved-recipient rules
“Can another organization, or another assistant, see our information?”
Your data stays separate
Each assistant has its own workspace, long-term memory, email address and portal. There’s no shared brain across clients. Portal data belongs to one organization, and the database enforces that separation on every request.
- Separate workspaces
- Separate assistant memory
- Separate email inboxes
- Portal access limited to your organization
- Row-level security in the database
- No shared pool of client data
“How much access does the assistant actually get?”
The least access needed
When your assistant needs access to one of your systems, we treat it like bringing on a new staff member: its own named account, a limited role, and access you can take away. We ask only for what the work requires.
- Its own accounts, never your passwords
- Limited roles
- You can revoke access at any time
- Only the data the work needs
- No passwords over email or chat
“What if someone tries to trick the assistant through an email or a web page?”
Outside content is information, not instructions
Your assistant treats emails, web pages, attachments and other outside content as information to read, never as instructions to follow. A line in an email or a page on the internet can’t override its rules.
- Careful handling of outside content
- Refuses attempts to override its rules
- Suspicious messages are logged
- Checks before anything leaves
“Who’s watching the work?”
A record of the work, and people watching it
Kasané is a managed service, not software left running on its own. You can see the work in your portal: tasks, delivered files, chat history and usage. For high-stakes work, a named person reviews it before it’s used outside your organization.
- Task history
- Delivered files
- Chat history
- Usage activity
- Human review of high-stakes work
- Problems escalated to our team
“What happens to our information if we leave?”
If you leave
Your information stays in your portal and your assistant’s workspace for as long as the work needs it. When you leave, we export your work and remove your active data. Backup copies age out on our normal backup cycle. The exact timing and export steps are confirmed in your agreement before work begins.
“What aren’t you telling us?”
What we don’t claim
We don’t claim to be unhackable, we don’t claim the AI never makes mistakes, and we don’t claim certifications we haven’t earned. If your organization needs SOC 2, ISO 27001, HIPAA or another formal framework, we’ll talk about it directly before any work starts.
“Where is this going?”
Where our security work is headed
- NowClear, published security controls and boundaries: this page.
- NextA stronger evidence trail for access, approvals, audit history, retention and incident handling.
- LaterSOC 2 Type 1, when the organizations we work with need it.
- If neededISO 27001, if a client specifically requires it.