SAFETY & COMPLIANCE

Safety-by-Design Framework

Safety is not a feature you bolt on after launch — it's an architectural principle that shapes every design decision from day one. For a kids VR game, safety is also your strongest competitive advantage. Parents choose safe games. Platforms promote safe games. Regulators leave safe games alone. This framework covers communication safety, moderation tools, anti-bullying systems, and age-appropriate design principles.

82%
Parents: Safety is #1 Concern
82% of parents of 8-13 year olds cite online safety as their primary concern when their child uses VR, ahead of screen time (67%) and spending (54%) — SuperData/Nielsen 2024
$0.15-0.40
Moderation Cost per 1K Users
Monthly cost for automated moderation (AI chat filter + report queue). Human moderation adds $1.50-3.00/1K users. Scale with automation, reserve humans for escalations.
<4 hours
Incident Response Target
Maximum time from report to action for severe incidents (threats, predatory behavior). Non-severe reports: 24-hour response target. Auto-action for critical keywords: instant.
+35%
Trust Score Impact on Retention
Games with visible safety features (report buttons, moderation badges) see 35% higher D30 retention from parent-approved players vs. games without visible safety UI

Communication Safety

Communication is the highest-risk surface in any multiplayer kids game. Every channel — text, voice, emotes — must be designed with the assumption that bad actors will attempt to misuse it.

Pre-Built Message Options (Quick Chat)

  • Default for all under-13 accounts. No free-text input. Players select from categorized message wheels.
  • Categories: Greetings ("Hi!", "Let's go!"), Gameplay ("Follow me", "Over here!", "Help!"), Reactions ("Nice!", "Wow!", "Oops!"), Social ("Want to be friends?", "Good game!", "See you later!")
  • Contextual messages: Show different options based on game state. During an escape: "Look at this!", "I found a clue!", "Hurry!". In lobby: "Ready?", "Wait for me", "Which room?"
  • Custom quick-chat: At Level 15+, players can create custom quick-chat messages from an approved word list. Pre-screened by content filter. This gives the feeling of personalization without free-text risk.
  • Emotes: 15-20 animated emotes (wave, dance, thumbs up, celebrate). No emotes that can be used to harass (no rude gestures, no "crying" emote). Emotes are visible to nearby players only.

Filtered Text Chat (Opt-In)

  • Requires parental opt-in through Meta Family Center or in-app VPC
  • Real-time filter checks every message against: profanity dictionary (10K+ terms including slang, abbreviations, leet-speak), PII patterns (phone numbers, addresses, email formats, social media handles), predatory language patterns (grooming phrases, meeting requests)
  • Blocked messages are replaced with "[message filtered]" — the sender sees a warning, the recipient sees nothing
  • Message history retained for 72 hours for moderation review, then deleted (COPPA data minimization)
  • Character limit: 80 characters per message. No links, no images, no file attachments

Voice Chat Policy

  • Disabled by default for all under-13 accounts. This is non-negotiable for COPPA compliance and child safety.
  • If parent enables voice: Apply real-time voice modulation (robot voice, animal voice) so children cannot be identified by their voice. This also makes the experience more fun and age-appropriate.
  • Voice toxicity detection: AI-based system monitors for aggressive tone, profanity, and suspicious conversation patterns. Auto-mutes flagged users and creates a report.
  • Mute controls: One-tap mute any player. Mute-all option always visible. Auto-mute players who have been reported by 2+ players in the session.
  • Party-only voice: Even when enabled, voice chat works only within friend parties, not with random matchmade players. This dramatically reduces risk.

Moderation Tools Matrix

Tool Purpose Auto / Manual Complexity Effectiveness
Word/Phrase Filters Block profanity, slurs, PII, and inappropriate language in real-time Automated Low — dictionary-based with regex patterns High for known terms (95%+). Moderate for creative evasion (~70%). Must be updated monthly with new slang.
AI Chat Moderation Detect context-based toxicity, grooming patterns, and intent that keyword filters miss Automated Medium — requires ML model training on kids' chat patterns Very high (90%+ for trained models). Catches obfuscated language, context-dependent threats. Use services like Spectrum Labs, Two Hat, or OpenAI Moderation API.
Player Reporting Allow players to flag inappropriate behavior, chat, or usernames Manual (player-initiated) Low — UI button + report form + review queue Essential complement to automated systems. Kids are unreliable reporters (over-report friends as jokes, under-report real issues). Weight reports by reporter history.
Behavior Scoring Track player behavior over time — reports received, reports filed, mute frequency, block frequency — to build a trust score Automated Medium — requires data pipeline and scoring algorithm High for identifying repeat offenders. Low-trust players get matched with each other (trust-based matchmaking) or receive increased monitoring.
Shadow Banning Restrict disruptive players without them knowing — they can play but are invisible to others Manual (moderation team decision) Medium — requires parallel session handling Moderate — effective for reducing disruption without triggering ban-evasion behavior. Use sparingly and for limited duration (24-72 hours).
Temporary Bans Remove players from the game for a defined period (1 hour, 24 hours, 7 days) Automated (thresholds) + Manual (escalation) Low — account flag with expiration High for deterrence. Escalating ban ladder: warning → 1hr → 24hr → 7-day → permanent. Always explain the reason to the player and the parent.

Content Moderation for UGC

If Eggscape allows any user-generated content (custom room names, avatar designs, clan logos, shared replays), every piece of content is a potential safety risk.

UGC Moderation Pipeline

  • Pre-approval queue: All UGC goes through an automated filter first (text: profanity/PII filter; images: AI image classification for inappropriate content). Content that passes automatically publishes. Content flagged by AI goes to human review.
  • Community flagging: Players can report inappropriate UGC. Reports from trusted players (high behavior score) are weighted more heavily. 3+ reports auto-removes content pending review.
  • AI image scanning: Use computer vision models (e.g., Google Cloud Vision SafeSearch, AWS Rekognition) to detect inappropriate imagery in avatar customizations or shared screenshots. Flag content with nudity, violence, or hate symbols.
  • Limited UGC scope: The safest approach is to constrain UGC to predefined options. Instead of free-draw avatar creation, offer modular pieces (body type + color + accessories). This eliminates 95% of UGC moderation needs.
  • Creator accountability: Track which player created each piece of UGC. Repeated violations result in UGC creation privileges being revoked before a game ban is applied.

Anti-Bullying Systems

Bullying in kids VR is uniquely harmful because of VR's immersive presence — being verbally attacked or physically crowded in VR feels more threatening than the same behavior on a flat screen. Eggscape must address this proactively.

Detection Systems

  • Repeated negative interactions: Track when the same player targets another player across multiple sessions (blocking their path, following them, emote-spamming at them). Flag after 3+ incidents between the same pair.
  • Personal space violation: In VR, players can invade others' personal space. Implement a "personal bubble" — when another player gets too close, their avatar becomes transparent and non-collidable. This is a game mechanic, not a moderation tool, making it feel natural.
  • Group targeting: Detect when multiple players coordinate against one player (all reporting a single player falsely, or all crowding one player). Flag coordinated behavior separately from individual incidents.

Intervention Mechanics

  • Instant escape: A "Safe Zone" button always accessible from the wrist menu. Teleports the player to a private space where no other player can reach them. They can rejoin when ready, report the incident, or leave the session.
  • Auto-separation: If the behavior scoring system detects a conflict between two players, auto-assign them to different teams/rooms in future matchmaking. No notification — it just happens.
  • Cool-down lobbies: If a session becomes heated (multiple reports filed), gently end the session for all players with a "Time for a break!" message and a 5-minute cooldown before re-matchmaking.

Victim Support

  • After a report, show the reporting player: "Thanks for telling us. We take this seriously and will review it." Provide immediate relief (the reported player is auto-muted and will not appear in their next 5 sessions).
  • Follow up with the player in-game next session: "We reviewed your report. We've taken action." (Don't disclose specifics about punishment.)
  • Parental notification: If a child reports bullying, notify the parent through Meta Family Center so they're aware and can have a conversation.

Bully Reform Paths

  • First offense: Warning message explaining what they did and why it's not okay. Written in kid-friendly language, not legalese.
  • Second offense: 24-hour ban + parental notification. Parent receives a message explaining the behavior and the consequence.
  • Third offense: 7-day ban. Player must complete a brief "digital citizenship" activity (age-appropriate quiz about being kind online) before returning.
  • Continued offenses: Permanent ban from social features (can play solo only) or permanent account suspension.

Age-Appropriate Design Principles

The UK's Age Appropriate Design Code (Children's Code) and the proposed US Kids Online Safety Act establish principles for designing digital products for children. Even if not legally binding in your jurisdiction, following these principles demonstrates due diligence and builds trust.

Design Standards for Eggscape

  • No dark patterns: No fake "X" buttons on promotions. No countdown timers that pressure purchases. No "spinning wheel" reward mechanics that mimic gambling. No making the "buy" button bigger/brighter than "skip."
  • Clear consent: When asking for anything (notification permissions, friend requests, data collection), explain in simple language what the child is agreeing to. "Can we send you a message when your friends are playing?" not "Enable push notifications for enhanced social experiences."
  • Understandable language: All in-game text should be readable by an 8-year-old (3rd grade reading level). Use short sentences. Avoid jargon. Test readability with actual children during UX research.
  • No manipulative mechanics: No "energy systems" that deplete and require payment to refill. No "pay or wait" timers. No artificial scarcity on functional items (cosmetic scarcity is acceptable). No streak penalties that punish children for not playing daily.
  • Positive reinforcement: Celebrate achievements, don't punish failures. "You almost made it! Try again?" not "You failed. 2 lives remaining." Frame everything around growth and encouragement.
  • Play time awareness: After 45 minutes, show a gentle "You've been playing for a while! Take a break and stretch." message. After 90 minutes, increase frequency. Never force a stop (that's the parent's role), but nudge awareness. This also satisfies platform requirements and builds parent trust.

Safety Implementation Tiers

Phase safety features across three tiers to balance development resources with risk coverage. Launch with Basic (non-negotiable), build to Standard within 3 months, and achieve Advanced by 6 months.

Feature Basic (Launch) Standard (3 Months) Advanced (6+ Months)
Chat System Quick-chat only (pre-built messages). No free text. No voice. Filtered text chat with parental opt-in. AI moderation layer. Voice chat for friend parties with voice modulation and AI toxicity detection.
Content Filtering Keyword/phrase filter on usernames and display text. 5K+ term dictionary. AI-powered contextual moderation. PII detection. Grooming pattern detection. Multilingual filtering. Image/UGC scanning. Behavioral pattern analysis across sessions.
Reporting One-tap report button. Pre-categorized reasons. Manual review queue (24hr target). AI-assisted report triage (priority scoring). 4-hour target for severe reports. Follow-up notifications. Automated action on high-confidence reports. Behavior scoring influences matchmaking. Full audit trail.
Player Protection Block/mute players. Personal space bubble. Safe Zone teleport button. Auto-separation of conflicting players. Cool-down lobbies. Parental incident notifications. Trust-based matchmaking. Predictive intervention (flag at-risk interactions before reports). Bully reform program.
Parental Tools Meta Family Center integration. Play-time visibility. Purchase notifications. In-app parental dashboard. Friend list visibility. Activity summaries. Feature toggle controls. Weekly safety reports to parents. Customizable safety settings. Direct communication channel for parent concerns.
Data Protection COPPA-compliant data minimization. No third-party tracking SDKs. 72hr chat log retention. On-device analytics processing. Session-scoped identifiers. Automated data deletion pipeline. Differential privacy for all analytics. Regular third-party privacy audits. Transparent data practices report.

Safety as a Marketing Advantage

Most kids VR games treat safety as a cost center — something you do because regulators require it. Eggscape should treat safety as a differentiator that drives acquisition, retention, and revenue.

  • Parent-facing messaging: "Eggscape is built for kids, by people who care about kids." Feature your safety measures prominently on your website, store listing, and marketing materials. Parents actively search for "safe VR games for kids" — be the answer.
  • Store listing advantage: Include safety features in your Quest Store description. "Quick-chat only communication, parental controls, no ads, transparent pricing." This converts browsing parents who are comparing options.
  • Press and influencer angle: Pitch your safety-by-design approach to parenting bloggers, tech journalists covering kids' safety, and VR publications. "How one VR game is putting kids' safety first" is a story editors want to tell.
  • Competitive moat: Gorilla Tag's #1 complaint from parents is the toxic voice chat and bullying. Rec Room faces similar criticism. Eggscape can position directly against this: "All the fun of social VR, none of the worry." This is a genuine competitive advantage that's hard for incumbents to replicate because their existing player base would resist safety restrictions.
  • Platform favor: Meta is under regulatory pressure to prove Quest is safe for children. Apps that demonstrate best-in-class safety practices are more likely to receive platform promotion (featured placement, editorial coverage, event inclusion). Being the "model citizen" kids app is strategically valuable.
  • Retention through trust: When parents trust a game, they don't uninstall it during periodic "screen time audits." Games that parents feel good about survive the periodic purges. This shows up as +35% D30 retention for safety-forward titles.