Aeliana logo
AELIANA
Living technology. Digital soul. Hyper-human.
Log in Register Forgot password

Safety: Self-Harm Referral Protocol Summary

We maintain a self-harm referral protocol and publish this summary at aeliana.ai/safety (or successor URL). This page explains how crisis referrals may be displayed and what resources are provided.

1) Detection Scope

  • Automated classifiers and rule-based checks evaluate user text for potential self-harm, suicide, or acute crisis signals.
  • Signals may include ideation, intent, planning language, and repeated high-risk escalation patterns.
  • Detection is risk-reduction, not a perfect guarantee. False positives/negatives can occur.

2) Referral Flow

  • If risk is detected, we prioritize safety messaging over normal companion flow.
  • Users are shown immediate referral options (for example U.S. 988 and local emergency pathways).
  • When needed, additional interruption prompts can encourage pause/end and real-world help.

3) Content Boundaries in Safety Mode

  • The system refuses instructions or escalation content for self-harm.
  • Dialogue redirects toward support, grounding, and professional/emergency resources.
  • Companion tone may remain empathetic while staying policy-bound.

4) Logging and Compliance

  • We record safety-event metadata needed for abuse prevention, policy audits, and legal compliance.
  • Logs are handled under Privacy Notice controls and retained only as necessary for operational/legal windows.
  • This summary may be updated as law, policy, and safety systems evolve.

Crisis Resources (U.S.)

Safety and policy contact: admin@aeliana.ai