AI notice

Charged integrates an intelligent conversational assistant named One (“Charged AI Agent”), engineered to assist Discord server administrators with configuring auto-responders, modifying command prefixes, and querying documentation through natural language. This AI Notice discloses how our AI technology operates, its strict boundaries, and our privacy commitments.

1 – Architecture & tool execution

One is powered by Google's enterprise Gemini API models combined with Charged's authenticated server-side tool execution pipeline:

1. When an administrator asks One to perform an action (such as creating an auto-responder or changing a prefix), the AI parses the intent and invokes the corresponding isolated tool function.
2. Tools execute with strict server permission validation against our Supabase PostgreSQL database, ensuring updates only occur on servers you are authorized to manage.

2 – Strict ecosystem boundaries

The AI Assistant is strictly confined to the Charged ecosystem:

1. Charged Domain Only: One assists exclusively with Charged Discord bot capabilities, server moderation, auto-responder rules, and documentation.
2. Rejection of Out-of-Scope Inquiries: The assistant refuses general programming requests, academic homework, political topics, or unrelated tasks.
3. Anti-Manipulation Guardrails: System prompt injection attempts or requests to bypass server security boundaries are strictly rejected.

3 – Zero model training on user data

Your conversations and server configurations are NEVER used to train, retrain, or improve generalized public AI models. Interactions sent via the Gemini API are processed in accordance with Google Cloud enterprise privacy standards with zero training retention.

4 – Human control & auditability

Every change initiated by the AI Assistant is executed with full visibility. The dashboard automatically illuminates modified rules with real-time visual highlighting, and administrators retain complete authority to modify or delete any rule at any time.