PLAYER DRIVENACADEMY
← All pillars·ACADEMY PILLAR

Trust & Safety

Moderation, policy, anti-abuse.

── THREE LEVELS
How levels work →
L01Free · open

Foundations

Understand what Trust & Safety owns: vocabulary, key metrics, how this discipline breaks, signals to invest. Plus a short assessment.

Read the primer →
L02Beta · free

Operator

Run Trust & Safety day-to-day. Tactical sessions, real briefs from operators in the trenches, peer review. Studio brief required.

Explore Operator packs →
L03Live Cohort

Lead

Lead Trust & Safety at studio scale. Live cohort, instructor-graded artifacts, capstone brief, credentialed.

Cohort 01 · July 2026
── L01 FOUNDATIONS PRIMER

What Trust & Safety owns.

Trust & Safety used to mean a 'mute' button in the chat client. Then players started killing themselves over harassment. Then governments started writing laws (DSA in Europe, KOSA in the US). Then advertisers and platform holders started auditing your moderation. Now T&S is the difference between a game that grows and one that's pulled from app stores.

The discipline covers three things that used to live in three different orgs: real-time moderation (in-game chat, voice, UGC), policy work (terms of service, age gating, behavioral expectations), and crisis response (the discord raid, the doxxing incident, the regulatory letter). The best T&S teams have one strategic owner across all three.

Trust & Safety isn't a cost center. It's the platform that lets everything else stand up.
Recurring theme on Player Driven

If you're targeting under-18s, you don't have a choice — KOSA, COPPA, the UK Age-Appropriate Design Code all carry real teeth now. If you're shipping in voice, the moderation problem is 10× harder and 100× more expensive. Studios that didn't take this seriously two years ago are scrambling now.

── THE OPERATOR'S CHEAT SHEET

Everything Trust & Safety owns, on one screen.

↳ WHAT YOU MEASURE
  • Reports per 1k DAU
  • Time-to-action on policy violations (median, p99)
  • False-positive rate on automated moderation
  • % of suspended accounts that re-offend after return
  • Regulatory compliance rate (DSA, COPPA, etc.)
↳ WHO OWNS THIS

Trust & Safety Lead, ideally reporting to a C-level not buried under marketing. At scale: separate Policy, Operations (moderators), and Engineering (tools + ML) functions.

↳ SIGNALS YOU NEED TO INVEST
  • You're shipping voice chat
  • You're targeting players under 18
  • Your moderator-to-DAU ratio is below industry norms (1:50k is bad, 1:10k is OK)
  • You haven't done a policy review in 12+ months
  • A regulator (Apple, Google, EU, FTC, ESRB) just sent you a letter
↳ HOW IT BREAKS — COMMON MISTAKES
  • Outsourcing moderation entirely to a vendor and forgetting to read the reports
  • Treating T&S as a privacy problem only (it's a safety problem)
  • Writing policy in legalese players can't parse
  • No appeals process — every false-positive ban is a future review-bomb
  • Forgetting that the moderators themselves need wellbeing support; vicarious trauma is real
── HOW GAMES USED THIS

Three studios. Three lessons.

Riot · Behavioral systems 2012–present

Riot Games (League + Valorant)

Riot's behavioral team pioneered the Tribunal (player-juried punishment), then ML-driven detection that catches toxic behavior before reports come in. They publish quarterly transparency reports. T&S is treated as a publishable competitive advantage.

Transparency on what you punish — and what you don't — turns moderation from a black box into a trust signal.

Roblox Corp · 2006–present

Roblox

Built one of the largest T&S operations in any tech company, period — thousands of moderators, ML scanning every uploaded asset and chat message. Has had public failures and recoveries; the size of the operation reflects the size of the problem they took on.

If your platform invites user-generated content, the moderation budget is the platform.

VRChat Inc · Voice moderation evolution

VRChat

Voice-only social VR is the hardest moderation problem in gaming. VRChat's path — community moderators, then audio fingerprinting, then ML voice classification — is the case study for what's possible when you can't read what people are saying.

Voice is where T&S goes when text moderation gets easy. Plan budget accordingly.

Reading is free and open. Create a free account to track your progress, earn credit for what you finish, and unlock Operator packs.

── RECOMMENDED READING

Latest from the network.

All in Trust & Safety