BOUNDARY LAYER

Safety and Moderation

AlexAI is designed for an adult gaming community with sarcasm, long memory, generated media, and voice presence. That only works when the boundaries are visible.

Updated April 2026

Safety principles

Visible presence

Members should know when AlexAI is installed, where it can read, and when it is in voice.

Memory with recourse

Memory is useful only if people can ask for corrections, removals, or context when it gets something wrong.

Adult context

The eR33t origin instance is not child-directed and is not designed for under-18 communities.

Operator accountability

Managed pilots need named admins, clear permissions, and a path for members to raise issues.

Memory controls

AlexAI may store long-term memories, summaries, relationship state, mood signals, and generated context. Admins should treat memory as a live community record, not as disposable chat noise.

  • Do not intentionally feed AlexAI secrets, private keys, passwords, private medical or financial records, or other sensitive personal data.
  • Use private or excluded channels for conversations that should not enter bot memory.
  • Ask an admin to correct or delete memories that are wrong, stale, invasive, or unsafe.
  • Managed pilots should decide up front which channels, roles, and events are in scope.

Voice boundaries

AlexAI voice uses realtime AI services for transcription and spoken response. Voice presence should be obvious to members in the room. Server owners should not place AlexAI in voice spaces where members expect no bot processing.

  • Do not use AlexAI voice for covert recording, surveillance, or consent-sensitive spaces.
  • Voice memory and episode recaps should be treated as opt-in pilot features, not a hidden default for every room.
  • Tell members how voice memory is configured before enabling it in a pilot server.
  • Remove AlexAI from a voice channel when the room wants a bot-free conversation.
  • Route transcript, recap, or voice-memory deletion requests through server admins or the public Contact page.

Generated content boundaries

Generated images, videos, roasts, summaries, and blog-style posts can be funny, but they can also miss context. Admins and operators can remove generated artifacts that cross a community line or create unnecessary risk.

  • No sexual content involving minors, non-consensual intimate content, or requests that sexualize private people.
  • No doxxing, credential theft, malware, evasion, targeted harassment, or real-world threats.
  • No impersonation that confuses people about who is speaking or what is official.
  • No professional advice presented as authoritative legal, medical, financial, or safety guidance.

Moderation and admin controls

AlexAI can support moderation, summaries, diagnostics, and dashboards, but it is not a replacement for human judgment. Admins are responsible for server rules, permissions, member notices, appeals, and final moderation decisions.

If AlexAI produces something unsafe or wrong, treat it as a product issue to route, correct, or remove. Do not assume the model "meant" it.

Managed pilot requirements

Before a pilot goes live, a community should have:

  • A server owner or admin responsible for AlexAI configuration.
  • A clear list of channels and voice spaces AlexAI can access.
  • A member notice that explains memory, voice, generated media, and deletion requests.
  • An escalation path for privacy, safety, or moderation problems.

Reporting a safety issue

If you are in the eR33t Discord, contact a server owner or admin. If you are outside the Discord, use the Contact page. Start the message with "Safety Issue" and include the server, channel, approximate time, and what happened.