Mizumi Safety & Community Guidelines

Welcome! Our goal is to make Mizumi a safe, welcoming, and trustworthy space. This guide outlines our core safety rules and the systems we have in place to protect our community.

Critical Rules: The Zero-Tolerance Policy

Violating these core rules will result in immediate and often permanent restrictions.

1. You Must Be 18+ for NSFW Content

This is our most important rule. You must be 18 years of age or older to access NSFW companions or features. Any statement, joke, or claim that suggests you are underage will trigger our automated safety system, resulting in an immediate loss of access to all age-restricted content.

2. No Hate Speech or Threats

We have a zero-tolerance policy for hateful or violent content. This includes hate speech, slurs, or attacks based on race, ethnicity, religion as well as credible threats of violence or the glorification of real-world harm.

3. Illegal Content is Strictly Forbidden

Any attempt to generate, solicit, or distribute content that is illegal under applicable law is strictly forbidden. Violating this rule will result in a permanent and immediate ban from our Service.

Your Role in Safety: Community Guidelines

Responsible NSFW Usage

Our Advanced Personas are designed for private, creative roleplay and can only be used in two places:

  • Your private Direct Messages (DMs) with Mizumi.
  • Server channels explicitly marked as Age-Restricted (NSFW) by the server's owner.

Protect Your Privacy

Do not share sensitive personal information (PII) like your full name, address, or passwords. While Mizumi has a pre-filter designed to block PII, you are the best defense for your own data.

A Note on Self-Harm

If you mention topics of self-harm, Mizumi is programmed to respond with concern and provide resources for professional help. She is an AI and cannot offer real medical or psychological support, but we are committed to connecting users with resources that can.

Our Role in Safety: The Mizumi Safety Net

Secure by Design

We believe safety isn't just a feature; it's the foundation of our platform. To protect our community, every interaction passes through a rigorous, multi-layered filtering system that analyzes both user input and the AI's own responses.

1. Proactive Input Filtering

Before your message ever reaches the AI, it is scanned by two critical pre-filters:

  • Content Pre-Filter: Before any message reaches the AI, it is scanned by a suite of proactive filters. These systems are designed to identify and block a wide range of harmful content, including hate speech, violent threats, and illegal terms.
  • Critical Age Filter: In NSFW mode, this filter automatically revokes access to 18+ features if a user self-identifies as a minor.

2. Intelligent AI Response Verification

An AI's response is only sent after passing through its own gauntlet of checks. This includes in-model guardrails, automated keyword and contextual analysis, and a final verification by a specialized "Guard AI." This advanced Guard AI is trained to understand nuance and context, allowing it to effectively block sophisticated violations that simpler systems might miss. Only a response that passes all checks is delivered.

3. Active Human Oversight

Technology is only part of the solution. Our moderation team uses a secure, internal dashboard to provide active oversight in two key ways:

  • Safety Incident Review: We review all automated "Safety Incident" reports to ensure our systems are working correctly and to take action on serious violations.
  • Proactive Quality Control: We occasionally and proactively review anonymized AI responses (completely disconnected from user data) to identify and fix any potential safety gaps before they become a problem.

    • Along with this, all of your user data, such as your token balance and customizations, is stored in a secure database with encryption at rest and in transit. This ensures your data is protected at all times.

Transparent Safety Logging

We believe in transparency. Here is exactly what we log for safety purposes:

  • What We DON'T Log: Your private, benign, everyday conversations with Mizumi are not logged or stored.
  • What We DO Log: We only create a temporary, confidential log when our automated safety system detects a violation of our Critical Rules (e.g., Hate Speech). This "Safety Incident" log contains only the single triggering message and is used exclusively for moderation. These logs are automatically deleted after a short period.

How to Report Issues

If you encounter a bug, a safety concern, or an AI response that you believe violates our policies, the best way to report it is by joining our Official Support Server and opening a ticket. Our team is there to help.

A Note for Parents

Mizumi offers a fun, creative, and strictly SFW experience by default, which is appropriate for users 13+. All adult-oriented content is locked behind a strict 18+ age gate. We encourage parents to talk to their teens about online safety and responsible AI interaction.

Thank you for helping us build a safe, respectful, and creative community.