Security and data handling
This page is a plain-language summary, not the legal text. See Privacy and Terms for that.
A human reviews replies by default
Out of the box, every reply the assistant drafts is just that, a draft. A person on your team reads it and clicks Send. Nothing goes to a customer unattended unless you deliberately turn on auto-send (email) or auto mode (the chat widget), and even then only for the specific replies that pass a fixed set of safety checks, see Auto-send. If you never turn those on, SupportWunder is a human-in-the-loop drafting tool, full stop.
The assistant only knows what you tell it
Drafts are generated only from your own knowledge base. Never from general internet knowledge, and never invented. If something isn't in there, the assistant says so instead of guessing.
The assistant is also guarded against passing along links, contact details, or claims it can't back up from what you've actually told it, so a customer message trying to trick it into repeating an unverified link or made-up contact doesn't make it into a reply unnoticed. This is a defense-in-depth safeguard, not a substitute for a human reading the draft.
Data handling and retention
Conversations, drafts, and the knowledge base you provide are stored so the product can work. Matching incoming messages, drafting replies, and showing you history. Reply generation is AI-assisted through a processor safeguarded by EU data-transfer protections. For the full details on what's collected, how long it's kept, which processors are involved, and your rights over it, see the Privacy policy. Contractual terms of use are covered in the Terms of service.
This page is a plain-language overview and isn't a substitute for the Privacy policy or Terms. If the two ever seem to disagree, the legal pages are authoritative.
Questions
For anything not covered here or on the legal pages, see Troubleshooting for how to reach support.