Moderation with Mustard Plays | Let's Chat

Overview

The video is a live, Q&A-style session on Fortnite Creative moderation, led by Epic executives from the Trust & Safety organization and related ecosystem leadership. The presenters repeatedly emphasize that moderation is a mix of automation and human review, with the goal of predictability and clarity for creators. They also note that enforcement exists but is often not publicly visible.

Moderation Workflow: Automation + Mandatory Human Review

• Human review for every published island

  • Epic states that every island published to Fortnite is reviewed by humans before it goes live.

• Automation to prioritize what humans inspect

  • Automated tooling helps route cases to human reviewers because there are thousands of island submissions/updates daily.

• Human gameplay checks

  • Reviewers typically play through islands (about ~10–15 minutes) to verify correct functionality and compliance with creator rules.

• Custom moderation tooling (“Badger”)

  • Epic uses a custom tool mentioned as “Badger” to surface questionable assets or parts of an island for deeper review.

• Context-aware moderation

  • Moderators must consider how elements are used within the island, not just whether assets are harmless in isolation.

Scale and Turnaround Expectations

• Volume

  • Epic reports about ~4,000 islands per day, many of which are updates that often review faster than brand-new submissions.

• Target review time

  • The stated operational goal for UEFN island moderation is that reviews generally finish within 3 hours, with exceptions depending on submission volume and seasonality.

• Staffing transparency

  • Epic avoids providing exact moderator staffing counts because staffing fluctuates with submission volume.

Consistency Problems (and How Epic Tries to Reduce Them)

Epic acknowledges that subjective judgment can cause variation (e.g., creators reporting thumbnails/assets being accepted multiple times and later rejected).

To reduce inconsistency, Epic highlights:

• Historical moderation decisions

  • If an asset has a strong prior approval record and nothing materially changed, it should require less re-review.

• Escalation / re-review on disagreement

  • If a decision conflicts with prior outcomes, the case is escalated so one moderator doesn’t “finalize” a rejection that contradicts earlier decisions.

• Training and monitoring

  • Tool-based reference guides, definitions, and examples support reviewers.
  • Epic also monitors performance and retrains when patterns drift.

• Gray-area challenges

  • Epic frames certain “gray areas” (like misleading player call-to-actions) as difficult to fully standardize through written rules alone.

Clearer Creator-Facing Communication

A major theme is that creators dislike “mystery” rejection messages. Epic says improvements include:

• Better rejection reasons
• More screenshots
• Plans for more structured guidance, such as comment-like explanations (e.g., describing what was too close or what triggered a tag)

Example “Gray Area” Discussed

• Misleading like/favorite call-to-actions
  • Epic explains that how close these are to enticing gameplay elements may not be captured well by a simple written rule (e.g., a distance threshold), which contributes to perceived unpredictability.

Enforcement and Sanctions

Epic indicates it enforces repeat violations (including a mentioned figure of roughly ~1300+ bans in the current year/period). A key clarification:

• Bans may not immediately remove already-published islands
  • Repeat-offender actions can block publishing and earning, but may not instantly remove live islands from players’ views.

Epic argues this is intentional to:

• Avoid leaving only punitive “public shaming” narratives
• Reduce “noise” while not disrupting players’ ongoing experiences unnecessarily

The presenters also discuss the tradeoff between:

• Removing violative islands to reduce unfair discovery competition
• Keeping player experience stable when an island already has an audience

Ratings and Age Classification (IARC)

• Epic does not author IARC ratings

  • IARC ratings are determined by a survey conducted by IARC, not by Epic.

• Epic moderation checks related to ratings

  • Epic performs checks that may flag inconsistencies (e.g., content not matching the submitted rating).

• Regional differences

  • Ratings can differ by region, and those variations do not necessarily indicate an error.

• Manual corrections by regional bodies

  • Epic notes that regional rating bodies can manually review and correct ratings, including audits to maintain consistency.

AI Stance: Compliance Over Creation Method

• Moderation is rules-based

  • Epic’s position is that moderation does not care what tool was used; it cares whether content breaks moderation rules.

• AI detection limitations

  • Epic argues AI is increasingly hard to detect reliably, so enforcement should not rely on “AI detection” alone.

• Broader AI philosophy

  • Embrace new technology, but emphasize ethical usage, licensing/ownership, and keeping humans central in the creative pipeline.

IP Violations: Permission Required + Partnership Programs

• Core policy

  • Don’t use other people’s IP without permission, or creators risk DMCA consequences.

• Proactive detection where rights-holder agreements exist

  • Epic references proactive IP detection when it has rights-holder agreements, including using supplied assets to train moderation systems and detect infringement upon upload.

• IP partnership program

  • Epic highlights growth of an IP partnership program to enable creators to use licensed content instead of infringing.

• DMCA limitations

  • Epic notes that rights-holders must file claims, so infringement can persist until rights are asserted.
  • Epic references improvements and enforcement mechanisms such as unoriginal content reporting.

Off-Platform Misleading Promotion (Actively Monitored)

Epic describes monitoring misleading off-platform promotion (YouTube/Twitch/TikTok), especially claims that promise rewards or mechanics that don’t exist in the island.

• Enforcement scale mentioned

  • Epic cites tens of thousands of DMCA takedowns of videos in the year.

• Attribution challenge

  • Epic can punish the on-platform creator only when it can reasonably determine the responsible party is linked to the misleading off-platform content, to avoid punishing innocent creators due to bad-faith third-party videos.

Roadmap: Prevention and Reduced Surprise

Epic closes with ongoing work to make moderation more predictable and less reactive:

• Prevent re-uploading duplicate thumbnails

  • Automatically detect originals already on the platform to reduce repeats.

• Pre-submit “look-ahead” warnings

  • Inside Creator Portal / UEFN, an ML classifier flags potential violations before publishing.

• Better handling of previously approved thumbnails/assets

  • Improvements aim to prevent later subjective reviewers from causing unexpected rejections without escalation logic.

• Moderator explanation comments

  • More detailed feedback to creators, intended to scale even though communication through writing has challenges.

Presenters / Contributors

• Sax Pearson (Epic executive team)

  • Oversees the Fortnite ecosystem, including UEFN and the creator economy.

• Dan Walsh (Lead product team for Trust & Safety)

  • Focuses on creator-facing moderation and enforcement systems.