Replace Community Sift with AI content moderation

Cinder will replace Community Sift for you. We train agents on your policies and the decisions your team has already made, test them on your own chat and content before anything changes, and move you over with no gap in coverage.

AI content moderation built around your policies

Cinder trains AI agents on your policies and past decisions to handle high-volume moderation and escalate the hard calls. Your community’s standards guide the decisions across chat, images, usernames, and other content.

Your team can inspect decisions, review escalations, and manage policies in one platform. Keep the tools you still need, with Cinder handling more of the work around them.

What you had with Community Sift, and how Cinder handles it

  • Topic classification

    Your policies, not a fixed taxonomy.

    Community Sift scored content against a set list of topics. Cinder agents are trained on your own policies and past decisions, so they draw the line where your team draws it. If you built your Sift setup around a custom taxonomy, we use it as the starting point.

  • User reputation

    Behavioral context on every decision.

    Player history and behavior go into every enforcement decision, so repeat offenders get caught even when no single message crosses the line.

  • Evasion detection

    AI moderation that adapts to evasion

    Players split words, swap characters, and move text into images to beat filters. Cinder agents are trained on the evasion your team has already caught, and we keep retraining them as players adapt.

  • Report triage

    Clear the obvious reports, escalate the hard calls.

    Agents work through the volume of player reports and send the ambiguous or severe ones to your team with context attached, so reviewers spend their time where judgment matters.

  • Per-feature policies

    A separate agent for every surface.

    Real-time chat, usernames, player profiles, in-game UGC, livestreams, and forums each need their own policy, latency, and context. Cinder trains the right agent for each surface and keeps policy, QA, escalation, and metrics in one system.

  • Child safety

    Governed workflows for the most sensitive cases.

    Cinder supports known-CSAM hash matching through established hash sources and approved reporting workflows. It also identifies child-safety signals such as grooming and solicitation, then routes those cases to trained reviewers.

Why gaming teams are using this moment to switch

Players don't wait for a migration.

Harassment, grooming, and evasion patterns keep evolving while you're between vendors. We run in parallel with your current setup, so nothing gets through during the switch.

The job got bigger.

Regulators are paying closer attention, players expect more, and your roadmap probably includes genAI features, group chat, and more UGC. Whether your org is called Player Support, Fan Care, or Trust & Safety, you need infrastructure that keeps up.

One team is accountable.

Cinder's agents and reviewers are run by one team that owns your accuracy, speed, and cost, so when something slips, one owner fixes it.

How the switch works

You take one call, and we handle the rest.

  1. We learn your setup.

    We go through your Community Sift configuration with you, including topics, sensitivity settings, languages, custom filters, username rules, and any automated actions tied to them.
  2. We prove it on your content.

    We train agents on your policies and past decisions, then run them in shadow mode against your real chat and content. You see the accuracy and speed for yourself, and Community Sift keeps running until you switch.
  3. We move you over.

    We switch your queues on an agreed date with no gap in coverage. Your policies, labels, and history come with you, and you keep any tools you're not ready to replace. Cinder can run headlessly, inside your existing queues, or in the Cinder platform.
3B+end users protected
94%of human review automated
3Xmore repeat-offender accounts taken down (Zello)
2Xmore abusive channels detected (Zello)

Frequently asked questions

Microsoft is winding down Community Sift, the moderation product originally built by Two Hat. Teams using it need a replacement for chat, image, and username moderation.

Talk to the team that moves moderation programs over for platforms like yours.