Naven
Glossary
Pre-send screening
Checking a message against a policy before it is delivered, rather than after somebody reports it.
Most moderation is reactive: a message arrives, somebody is hurt by it, somebody reports it, and hours or days later it is removed. The harm has already happened — removal is bookkeeping. Pre-send screening moves the check in front of delivery, so the message that would have caused the harm is never received.
The cost is that the server has to read the message, which rules out end-to-end encryption. The other cost is false positives: a screen that stops real harm also sometimes stops an ordinary sentence, and a system that pretends otherwise is not being straight about it.
Where Naven stands
This is the thing Naven does that its competitors structurally cannot. Twelve published categories, checked before send. It misses things, it is weaker in languages other than English, and it does not apply to calls at all — all three are stated on the page that explains it.