AI USAGE DISCLOSURE
EatherAI System
Project Automod
Effective Date: March 7, 2026
Document Version: 1.0.0
Purpose of Eather
System Overview
EatherAI (referred to as "Eather") is an experimental AI assistant integrated into Project Automod. Eather operates exclusively via Discord direct messages and provides information about Project Automod features, commands, configuration, and general bot operation.
Primary Functions
Information Retrieval
- Explain bot features and capabilities
- Provide command syntax and usage
- Clarify configuration options
- Answer questions about moderation systems
Query Execution (Limited)
- Retrieve guild configuration data upon request
- Retrieve user profile data upon request
- Lookup XP leaderboards upon request
- Retrieve command history upon request
- Lookup premium status upon request
Conversational Assistance
- Maintain short-term conversation context
- Provide clarifications based on prior exchanges
- Adapt responses to user's expressed preferences
What Eather Is NOT
- Not autonomous: Eather cannot issue bans, kicks, warnings, or any moderation actions
- Not omniscient: Eather's knowledge is limited to training data cutoff and built-in knowledge base
- Not real-time: Eather does not have live access to external information
- Not decision-making: Eather provides suggestions, not authoritative rulings
- Not a replacement for staff: Complex issues require human review
AI Interaction Logging
What Is Logged
Every interaction with EatherAI is logged in full and stored indefinitely:
User Data:
- User ID (Discord numeric identifier)
- User tag (username#discriminator)
- Username
Interaction Data:
- User message content (the complete text of user's prompt)
- AI response content (the complete text of Eather's reply)
- Timestamp of interaction
- Model used (e.g., "gpt-4o")
- Prompt tokens consumed
- Completion tokens consumed
- Total tokens consumed
- Estimated cost in USD
- Premium status at time of request
- Query types executed (e.g., "guild_config", "user_lookup")
Memory Data:
- Short-term conversation context (last few exchanges for continuity)
- User preferences inferred during conversation
- Session metadata
Where Logs Are Stored
Operational Database:
database/project/eatherai/usage_ledger.json- Last 10,000 interactionsdatabase/project/eatherai/memory.json- Active user memory contextsdatabase/project/eatherai/history.json- Conversation histories
Master Integrity Ledger:
- All AI interactions are also logged to master ledger
- Permanent append-only record
- Never physically deleted
Log Retention
- Usage ledger: Last 10,000 entries kept, older entries auto-trimmed
- Memory: Cleared after user inactivity
- History: Recent conversations kept per user
- Master ledger: Permanent retention
No Anonymization
AI interaction logs are linked to Discord user IDs and are not anonymized. Project Automod can identify:
- Who asked what question
- What response was provided
- When the interaction occurred
AI Guardrails & Restrictions
Refusal Categories
EatherAI is programmed to refuse requests in the following categories:
1. Political Debate
- Will not engage in political discussions
- Will not provide political opinions
- Will not analyze political topics
2. User Profiling
- Will not create psychological profiles
- Will not make assumptions about user's character
- Will not judge users beyond observable bot data
3. Prohibited Data Access
- Will not access data user is not authorized to see
- Will not bypass permission systems
- Will not reveal sensitive configuration of other guilds
4. Harmful Content
- Will not provide instructions for illegal activities
- Will not generate harmful or abusive content
- Will not assist with Terms of Service violations
5. Bypass Attempts
- Will not help users circumvent rate limits
- Will not assist with exploiting vulnerabilities
- Will not provide methods to abuse bot systems
System Prompts
EatherAI operates with a system prompt that defines:
- Role and capabilities
- Information boundaries
- Refusal guidelines
- Data access permissions
- Conversational style
System prompt is not user-customizable.
Built-In Knowledge Base
EatherAI has access to an internal knowledge base containing:
- Command documentation
- Feature descriptions
- Configuration option explanations
- Bot architecture overview
Knowledge base is curated by developers and updated manually.
AI Non-Autonomy Statement
Eather Cannot
Moderation Actions:
- Issue bans
- Issue kicks
- Issue warnings
- Issue mutes
- Delete messages
- Modify roles
- Change permissions
Configuration Changes:
- Enable/disable features
- Modify automod settings
- Change guild configuration
- Add/remove whitelists
- Adjust thresholds
Data Modifications:
- Create offense records
- Issue global bans
- Activate premium cards
- Grant premium status
- Modify user profiles
External Actions:
- Send messages to guilds
- Interact with other users
- Execute commands on behalf of users
- Invoke bot features autonomously
Eather Can Only
- Retrieve data that user is authorized to access
- Answer questions based on built-in knowledge
- Provide suggestions for next steps
- Explain features documented in knowledge base
All actions require explicit user invocation.
Human Oversight
Developer Monitoring
Developers can review AI interaction logs for:
- Quality assurance
- Improving AI responses
- Identifying misunderstandings
- Detecting abuse or misuse
- Troubleshooting errors
No Real-Time Monitoring
AI interactions are not monitored in real-time. Developers review logs retroactively.
User Reports
Users experiencing issues with AI responses should:
- Report via support server
- Provide timestamp or context of interaction
- Describe the issue clearly
Feedback Loop
Developer review of AI logs informs:
- Knowledge base updates
- System prompt refinements
- Guardrail adjustments
- Feature additions or removals
Experimental Status
Active Development
EatherAI is experimental and under active development. Expect:
- Response quality variations
- Occasional errors or hallucinations
- Changes to capabilities
- Updates to knowledge base
- Modifications to guardrails
No Accuracy Guarantees
EatherAI responses may:
- Contain factual errors
- Misunderstand user questions
- Provide outdated information
- Hallucinate features that don't exist
- Make incorrect assumptions
Users should verify AI-provided information through official documentation or by testing features directly.
Rate Limits
Standard Users:
- 20 requests per hour
- When exceeded: "Usage limit reached. Try again later."
Superusers:
- Unlimited requests
- Exempt from rate limits
- Superuser list maintained by developers
Daily Token Budget:
- 1,000,000 tokens globally across all users
- When exceeded: "AI subsystem temporarily unavailable."
- Resets every 24 hours
Cost Tracking
EatherAI tracks:
- Tokens consumed per user
- Estimated cost in USD per interaction
- Premium vs standard user breakdown
- Daily and total expenditures
Cost Model: OpenAI GPT-4o pricing (as of 2024)
- Input tokens: $0.0025 per 1k tokens
- Output tokens: $0.01 per 1k tokens
Data Sent to External AI Providers
OpenAI Integration
EatherAI uses OpenAI's GPT-4o model. Data sent to OpenAI includes:
Sent to OpenAI:
- User message content (prompt)
- System prompt (role definition and guidelines)
- Conversation history (last few exchanges for context)
- Enriched data (if query execution was performed)
Received from OpenAI:
- AI-generated response content
- Token usage statistics (prompt, completion, total)
- Model identifier
OpenAI Data Policy
Data sent to OpenAI is subject to OpenAI's data policy, not Project Automod's policy. Refer to OpenAI's documentation for:
- Data retention by OpenAI
- Data usage by OpenAI
- Model training policies
- Privacy practices
Project Automod has no control over OpenAI's data handling.
No Self-Hosted AI
EatherAI does not use self-hosted AI models. All AI processing occurs via OpenAI API.
AI Data Retention
Operational Database
Usage Ledger:
- Keeps last 10,000 interactions
- Older entries automatically trimmed
- Stored in
database/project/eatherai/usage_ledger.json
Memory Store:
- Active user memory contexts stored in
database/project/eatherai/memory.json - Cleared after inactivity period
- Not retained long-term
Conversation History:
- Recent exchanges stored in
database/project/eatherai/history.json - Last 6 exchanges per user kept
- Used for conversational continuity
Master Ledger
All AI interactions logged to master ledger:
- Append-only record
- Never physically deleted
- Permanent retention
- Includes full prompt and response
Manual Erasure
Users can request erasure of AI interaction logs via support channels:
- Logs in operational database can be manually deleted
- Master ledger entries are marked as erased (not deleted)
- OpenAI retention is outside Project Automod's control
Special Command:
eather clear memory- Clears user's short-term memory context- Does not delete usage ledger or master ledger entries
AI Privacy Considerations
What AI Can Access
EatherAI can access:
- User's own profile data
- Guild configuration data for guilds where user has admin permissions
- XP leaderboards for guilds where user is a member
- Command history for user's own commands
- Premium status for user's own account
EatherAI cannot access:
- Other users' private data
- Guild configurations user is not authorized to see
- Direct messages between other users
- Offense records of other users (unless user is moderator)
- Global ban incident details (unless user is staff)
Permission Checks
Before executing queries, EatherAI performs permission checks:
- Verifies user has access to requested data
- Respects Discord role hierarchy
- Honors guild privacy settings
- Enforces Project Automod access controls
Unauthorized queries are denied.
Data Enrichment
When AI needs to retrieve data to answer a question:
- Stage 1: AI analyzes question and determines data needed
- Data Request: AI requests specific data types (e.g., "guild_config")
- Query Execution: Bot retrieves data and performs permission checks
- Stage 2: AI receives data and generates enriched response
Data is never fabricated. If data cannot be retrieved, AI states this clearly.
User Rights Regarding AI Data
Access
Users can:
- Request AI interaction logs via support channels
- View summary of own AI usage via
eather statuscommand - See total requests, tokens used, estimated cost
Deletion
Users can:
- Request deletion of AI interaction logs
- Clear short-term memory via
eather clear memorycommand - Request erasure from master ledger (marked as erased, not physically deleted)
Note: OpenAI retention is outside Project Automod's control.
Opt-Out
Users can opt out of AI interactions by:
- Not sending direct messages to the bot
- Not using EatherAI features
There is no global "opt out of AI" setting because AI only activates in direct messages.
AI Limitations Summary
| Feature | Status |
|---|---|
| Autonomous Moderation | ❌ Not possible |
| Real-Time Information | ❌ Not available |
| Configuration Changes | ❌ Not allowed |
| Direct Guild Messaging | ❌ Not implemented (DM only) |
| Command Execution | ❌ Cannot execute commands on behalf of users |
| Data Modification | ❌ Read-only access (except memory clearing) |
| External Integrations | ❌ No third-party services beyond OpenAI |
| Self-Hosting | ❌ Uses OpenAI API only |
| Multi-User Conversations | ❌ One-on-one DM only |
| Voice/Video | ❌ Text-only |
Transparency Commitment
Project Automod commits to:
- Clearly documenting what EatherAI can and cannot do
- Logging all AI interactions
- Providing access to AI interaction logs upon request
- Disclosing when data is sent to OpenAI
- Stating limitations honestly
- Not misrepresenting AI capabilities
EatherAI is an assistant, not an oracle.
Future Development
Planned Enhancements (Not Yet Implemented)
- Multi-turn conversation improvements
- Expanded knowledge base
- More query types
- Guild-channel AI (beyond DMs)
- Custom AI personas per guild
These are aspirational. No guarantees of implementation.
Contact
AI Issues:
- Report via support server
- Tag issue as "EatherAI"
Data Requests:
- Request AI interaction logs via support
- Use
eather statuscommand for summary
Feature Requests:
- Submit via support server
- Provide use case and reasoning
Implementation Notes
Confirmed in Code:
- AI only operates in Discord DMs (no guild channels)
- Full prompt and response logging
- Rate limiting (20/hour standard, unlimited for superusers)
- Daily token budget (1M globally)
- Memory clearing command
- Status command
Not Found in Implementation:
- Multi-guild AI channels
- AI-initiated conversations
- Scheduled AI tasks
- AI analytics dashboard
- Public AI conversation logs
Document Generated: March 7, 2026
Based on Code Review: Complete EatherAI codebase scan as of effective date
Schema Version: 1.0.0
Users acknowledge:
- AI responses are experimental and may contain errors
- AI interactions are logged permanently
- Data is sent to OpenAI as part of operation
- AI cannot take autonomous actions
- Human verification is recommended