Iris — Privacy Policy
Last updated: August 5, 2026 Operated by: human.tech (Holonym Foundation)
Iris is an AI-powered community support bot that runs in the human.tech Discord server. This policy describes what data Iris accesses, stores, and how you can control it.
What Iris reads
Message content across the human.tech Discord server for two purposes:
-
AI-powered support — detects product questions and responds with answers grounded in a curated knowledge base of 675+ sources. If the knowledge base doesn’t cover a topic, Iris returns a standardized “I don’t know” response and escalates to a human moderator. Iris answers only from the knowledge base and does not draw on outside knowledge.
-
Link and scam protection — passively scans messages server-wide to detect and remove known phishing and scam URLs, protecting community members in real time.
Messages that do not trigger a bot action are not logged or stored.
What Iris stores
| Data | Where | Retention |
|---|---|---|
| Ticket transcripts (message text, Discord user ID, username, timestamps) | Private GitHub repository | Retained for knowledge base reference |
| AI interaction logs (question, response, sources used, model, token usage, channel ID, user ID) | Local JSONL files on application server | 90 days (auto-pruned) |
| Ticket event logs (open, close, escalate, category) | Local JSONL files | 90 days (auto-pruned) |
| Feedback and CSAT logs (thumbs up/down, survey responses) | Local JSONL files | 90 days (auto-pruned) |
| Scam filter logs (flagged content truncated to 500 chars, user ID, channel, matched domains, action) | Local JSONL files | 90 days (auto-pruned) |
Web-initiated tickets may also include email and wallet address if voluntarily provided by the user.
What Iris does NOT store
- Messages that do not trigger a bot action
- Bulk message logs
- User profiles or behavioral data
AI use
Messages from support interactions are sent to OpenRouter (openrouter.ai) for AI response generation — inference only, never used for model training. OpenRouter routes requests to:
- Google Gemini Flash — majority of standard support queries
- Anthropic Claude — default and audit-level responses
All inference requests are routed exclusively through OpenRouter. human.tech does not train, fine-tune, or build any machine learning model using message data.
Third-party data sharing
| Recipient | Purpose | Data shared |
|---|---|---|
| OpenRouter (→ Google Gemini, Anthropic Claude) | AI inference | Message content from support interactions |
| GitHub (private repository) | Ticket archival | Ticket transcripts |
No data is sold, shared for advertising, or used for model training. All other data is stored on the application server controlled by human.tech.
Retention
- Interaction logs, ticket event logs, feedback logs, CSAT logs, and scam filter logs are automatically pruned after 90 days. The retention sweep runs on startup and every 24 hours.
- Ticket transcript archives (markdown files on GitHub) are retained for knowledge base reference.
- No indefinite bulk storage of general messages occurs.
Opt-out
Iris provides multiple opt-out mechanisms:
!iris-optout— type in any channel to stop AI responses to your messages.!iris-optinreverses it.!iris-statusshows your current status.- Mute AI button — available in every ticket thread for both the ticket creator and moderators. Stops AI follow-ups in that ticket.
- Link/scam filtering cannot be opted out of — it is a server-wide safety feature, similar to Discord’s native AutoMod. No data is stored for messages that don’t match the blocklist.
Direct messages
Iris does not send direct messages. All interactions happen in-server.
Data deletion
To request deletion of your data, contact maylynne@holonym.id or message 0xblackmarq on Discord.
Upon request, all interaction logs, ticket records, and archived transcripts associated with your Discord user ID will be permanently removed.
Contact
maylynne@holonym.id | Discord: 0xblackmarq