AI & Internal Intelligence
AI is a guidance and experience layer, never a final authority. Members are always free to ignore AI output.
- ◆AI runs server-side only — provider keys never reach the browser.
- ◆No biometric AI, no facial recognition, no automated decisions with legal effect.
- ◆AI moderation assists human admins; final action is always human-reviewed.
- ◆Used for narration, translation, memory enhancement, and moderation triage.
- ◆Internal intelligence is observational, not autonomous.
Era status
◇An observational AI layer can earn trust during the founding era; autonomy never can.
Where AI is used today
Narration and ambient text — soft contextual lines surfaced on calm surfaces. Translation — per-message translation across diaspora languages. Memory enhancement — optional AI text, banner image, mood, and date interpretation when a member contributes a memory. Moderation triage — severity classification and tag extraction on anonymous community reports, surfaced to human admins for review. Chat bot participation — clearly labelled bot accounts that contribute to public threads.
What AI never does
No automated decisions with legal effect. No biometric processing. No facial recognition. No advertising profiling. No autonomous account actions. AI never sees personal cross-surface history — only the context required for the immediate task.
Data sent to model providers
Narration receives anonymised session context (time, device class, surface, prior opt-in tags). Translation receives a single message. Memory enhancement receives the member's memory body and chosen prompt. Moderation triage receives the submitted report text. Provider contracts forbid use of this content for training.
Disclosure
AI-generated content within member-facing surfaces is labelled or distinguishable. Bot participants are flagged. Members can opt out of optional AI features at any time through notification and surface settings.