The safest peer
platform ever built.

How it works

Every message is scanned by a fine-tuned NLP classifier before delivery to the recipient. The model checks for self-harm language, suicidal ideation, bullying, hate speech, sexual content, and personal identifying information exposure.

Flagged messages are held in a quarantine queue and reviewed by the human moderation team within 5 minutes. The sender sees a brief notification. The recipient never receives the flagged message.

Technical specs

• ModelGPT-4o mini + fine-tuned BERT
• Latency<200ms per message
• Accuracy>97% precision, >94% recall
• Categories8 harm categories monitored
• UpdatesRetrained monthly on new data

How it works

Every AI flag reaches a trained human moderator within 5 minutes during peak hours, and within 15 minutes at all other times. Moderators complete mental health first aid training and follow documented escalation protocols based on severity.

Moderators can: warn and release, remove content, temporarily suspend accounts, issue permanent bans, or escalate to a licensed mental health professional.

Moderator standards

• TrainingMental Health First Aid certified
• Coverage24/7 across all time zones
• Response (peak)<5 minutes
• Response (off-peak)<15 minutes
• ReviewMonthly audits + supervisor oversight

How it works

When imminent self-harm or suicidal intent is detected — by AI or by a human moderator — the conversation is immediately escalated to a licensed mental health professional on call. Simultaneously, crisis resources are surfaced in-app for both participants.

The peer user is never expected to handle a crisis situation alone. Our approach follows established safe messaging guidelines from the American Foundation for Suicide Prevention.

Crisis resources shown in-app

• US988 Suicide & Crisis Lifeline
• US (text)Crisis Text Line — text HOME to 741741
• UKSamaritans — 116 123
• AULifeline — 13 11 14
• CACrisis Services Canada — 1-833-456-4566

How it works

Block, report, and end-chat buttons are always prominently visible in the UI — never buried in settings, never hidden behind confirmation dialogs. Tapping block takes instant effect with no friction.

Reports go directly to the moderator queue with high priority. Users can optionally add context. Reported users are flagged for immediate monitoring and reviewed within 1 hour.

User control specs

• BlockInstant — no confirmation needed
• Report reviewWithin 1 hour, always
• Categories6 report types + free text
• OutcomeEmail notification within 24hrs
• AppealsAll bans are appealable

How it works

For users aged 13–15, Solaria requires parental or guardian consent during onboarding. Once completed, the guardian receives access to an optional dashboard showing activity summaries — never chat content.

Chat content privacy is non-negotiable. Research shows teens are far less likely to seek support if they believe they're being monitored. We protect that trust absolutely — while still giving parents meaningful visibility into usage patterns.

What parents can see

• ✓ VisibleSession count and duration
• ✓ VisibleForum participation overview
• ✓ VisibleUsage hour patterns
• ✓ ConfigurableUsage hour limits and restrictions
• ✗ NeverChat content — ever