HavenDOCS
Back to Home

Safety & Moderation (Haven 2026)

Haven maintains a high standard of safety for its 14+ audience through automated and multi-layered moderation systems.

🤖 AI-Powered NSFW Scanner

Haven uses a TensorFlow.js model (MobileNetV2) to perform decentralized, client-side moderation for 1-on-1 video chats.

How It Works:

  1. Sampling Frames: Every few seconds, the HiddenScannerNode in ChatCallProvider.tsx captures a 224x224 frame from the remoteStream.
  2. On-Device Inference: The frame is processed entirely in the visitor's browser. Images are never uploaded to a server for scanning, ensuring maximum privacy.
  3. Violation Callback: If a frame is scored as NSFW (>85% confidence), the system triggers:
    • Instant Disconnect: The call is terminated immediately.
    • Auto-Report: The reportedUserId and Automated AI Ban reason are sent to the database.
    • Matchmaking Blacklist: The WebSocket server adds the offender to a temporary memory-based ban list.

🛡️ Manual Reporting System

Users can manually report partners during or after a call.

  • Fast-Report Trigger: A report immediately ends the current session on the reporter's end.
  • Moderation Queue: Reports are persisted to the reports table in Supabase.
  • Reputation (Karma): Each report impacts the user's hidden reputation score, eventually leading to a permanent ban or shadow-ban in the matching pool.

🚪 Access Control & Verification

  1. Age Verification:

    • 1-on-1 video matching is strictly limited to users who have completed the age verification self-declaration (managed via AgeVerificationModal.tsx).
    • Verified status is cached for performance but periodically re-validated by the verification middleware.

  2. Email Confirmation:

    • Mandatory for all interactive features (Chat, Sangha creation, Friend requests).
    • This ensures a valid point of contact for moderation notices and reduces bot spam.

  3. Sangha Moderation:

    • Role-Based Control: Creators can define moderators with permission to kick or ban users from a specific Sangha.
    • Auditing: Session participants are tracked in room_participants for accountability.

💡 Safety Tips

  • Always Report Unusual Behavior: Don't just skip; reporting helps keep the community clean for everyone.
  • Status Badges: Look for the "Verified member" badge in Sangha lists for a safer experience.

Created by the Antigravity Safety Team.