Platform

Moderation that reaches every corner

One engine across every surface, AI that can hide but never punish, humans who decide, and an appeal for every decision. Built so two people can keep a professional community safe.

Every write goes through one engine

Posts, comments, profiles, avatars, listings, proposals, event descriptions, chat, course text and group text all pass through the same decision. Abuse moves to whatever is unwatched, so nothing is unwatched.

AllowThe normal case. Nothing happens.

QueueStored and visible; a human is asked to look.

HideStored but invisible until reviewed. Enforced in the database, not the UI.

BlockRefused at the door. Only for a small set of severe categories, only from a deterministic filter, never from a model guess.

AI hides. People decide.

A classifier runs on everything after it's written and can hide content for review. It can never issue a strike, suspend or ban. Every sanction is a human decision, which is what makes every sanction appealable.

Deterministic first passA fixed filter for the severe categories, synchronous, explainable.

Model second passClassifies after the write; capped at hide. Swappable provider.

Behaviour signalsTrust level and posting patterns change how closely something is looked at, never whether someone is punished.

Precision mattersIn a professional community a false positive costs someone a client. Overturns feed back as training signal.

Reports, queue, sanctions

Report from anywhereEvery surface carries a report control, including chat and profiles.

One queueModerators see everything in one place, sorted by severity and age.

A ladder you configureWarning, strike, suspension, ban, with thresholds you set and a member-facing policy that matches them.

Enforced at the databaseA suspended member can't post through the app, the API or any other path. Not a hidden button, a refused write.

Removal is quarantine

Moderators never hard-delete. Removed content leaves every read path but is retained, so a mistaken decision can be reversed and anything with a legal preservation duty is still there. Permanent deletion is a separate, privileged retention operation.

AppealsEvery decision can be contested. An upheld appeal reverses by writing a new decision; the history stays intact.

Audit trailWho did what, when, to which content, with the previous state.

No labels on peopleVerdicts attach to content. The engine never writes a risk flag or a health inference onto a member.

Self-harm and CSAMHandled outside the strike ladder: restricted visibility, immediate human escalation, content preserved, no automated message to the member, a vetted global resources link surfaced instead.

Your policy, your guidelines

Categories and thresholdsSet in the admin console, validated before they take effect.

Community guidelines pageWritten in your language, in your voice, linked from every report dialog.

Moderators you appointMembers, staff, or Beakforce operators if you'd rather we did it.