Transparent by design: what the other party is told, and why it matters

A practitioner weighing a message filter will sooner or later wonder how it looks from the other side, and how it might look to a court. Filtering someone’s messages in secret could be painted as surveillance or deception, and that is the last thing you want attached to your client. FenceChat was designed so that question doesn’t arise.

What the other party receives

The first time someone sends a message to your client’s FenceChat email address or number, FenceChat sends them one short notice, by email or text to match how they wrote. It tells them their messages are processed by AI and stored by FenceChat before delivery, that this is a one-time notice, and where to read our privacy policy.

It doesn’t name your client, accuse anyone of anything or describe what the filter removes.

The notice is part of how FenceChat meets its privacy obligations: Australian privacy law requires a business to tell people when it collects and stores information about them, including when that information comes from someone else. The short “Sent via FenceChat using AI” line on replies is there so it’s clear that AI was involved in the service.

What they don’t receive

• No further notices after the first one, apart from a short “Sent via FenceChat using AI” line on replies your client sends through FenceChat.
• No flag on any individual message saying it was filtered.
• No copy of the filtered version, and no list of what was removed.
• No request to sign up, download anything or pay.

They carry on emailing or texting as they always have, and they keep their own copies of what they sent. Their original words are also kept, unchanged, in your client’s secure and encrypted Archive.

Why we chose openness

A filter that worked in secret would have been easier to build. We chose openness deliberately, for four reasons.

1. Fairness to the sender. People are entitled to know, in general terms, that their messages pass through a service that stores them. A one-time notice does that without exposing your client or the content of anything.
2. Your client’s credibility. If communication is ever examined, it is hard for anyone to claim your client was secretly altering or concealing what the other party wrote. The original is kept, it’s able to be read by your client, and the other party was told.
3. What makes FenceChat different was never secrecy. It is that only your client has to set it up. The other party doesn’t need to agree, join or install anything, and that is what makes it work where shared apps stall.
4. It may encourage calmer messages. Knowing their messages are kept on record can lead some people to write more carefully. Research on post-separation parenting apps found that, for many parents, the app helped discourage abusive messages as well as record them. It won’t change everyone, and for some people, learning their messages are filtered can provoke anger, which is why the safety option exists. It’s no substitute for safety planning, but it is one more reason openness can help.

When telling them would be unsafe

For most clients, the one-time notice to the sender is simply the right thing to do. For some, especially where family violence is a concern, telling the other party anything at all could create risk. So your client can turn the notice off in their settings, with the reason recorded as a safety concern. That is an exception for safety, not a preference setting.

Texts from unknown numbers

FenceChat only relays texts from numbers your client has added as contacts. A text from any other number isn’t filtered, stored or passed on, so it doesn’t get the notice either. It gets a standard auto reply explaining that the number only receives SMS messages from contacts its user has added.