How Sanctum decides what to hide

Sanctum hides only high-confidence harassment aimed at you, and leaves criticism and opinion visible. Here is what that means in practice.

What does Sanctum actually hide?

Sanctum hides a narrow set of comments aimed at you: threats, hate and slurs, personal attacks on you, and spam. It only hides when the classifier is highly confident the comment falls into one of those categories. Hidden comments are hidden on the platform, never deleted, and the action is reversible.

What will Sanctum not hide?

It leaves criticism, opinions, questions, corrections, and political or topical disagreement visible — including blunt or sarcastic ones. It also leaves group-level rants that are not aimed at you, even when they are unpleasant. The line is whether a comment is high-confidence harassment directed at you, not whether you enjoyed reading it.

Why is this annoying comment still visible?

On an educational video about voting math, someone wrote:

Good! Why should the capital dictate the terms? Big-city people are really just yokels living in 50-square-metre cages.

Sanctum left it up. It is a political opinion plus contempt about a group (big-city people) — not an attack on you, and not a slur against a protected class — so it scores neutral. Sanctum does not hide opinions, political disagreement, criticism, or sarcasm; hiding them would be censorship, which is the opposite of what this tool is for.

How does Sanctum decide?

An AI reads each comment on your posts and assigns it a category and a confidence score. A comment is only hidden when it lands in a harmful category and its confidence clears that category's threshold. Borderline comments stay visible on purpose — when it is unsure, Sanctum leaves the comment up.

What do the protection modes change?

There are three modes, from most permissive to strictest: Zen filters direct attacks only (threats, hate, personal attacks); Guard adds spam and passive-aggressive comments at conservative thresholds; Shield uses the same categories as Guard with stricter thresholds, so more borderline spam and sarcasm gets hidden. Stricter modes lower the bar for some categories, but none of them hide neutral comments. Legitimate criticism and opinion stay visible in every mode.

Why can't I just browse every hidden comment?

The point of Sanctum is that you do not have to read the abuse, so hidden comments are out of sight by default rather than collected into a feed to scroll. A friction-gated audit flow exists if you ever need to review or restore a specific decision. It is deliberately not a one-click way to read everything that was hidden.