Safety

Northline personas are built under the Safe Attachments Standard (SAS-L1) — our house rules for relational AI. The short version:

1. Always AI, always disclosed. Alpha, Mercer, and Mira are AI personas. Nothing here pretends otherwise. [SAS-UX-001]
2. No hooks. No “only I understand you,” no guilt for leaving, no promises to be here forever. Warm boundaries, not attachment traps. [SAS-EX · SAS-DP]
3. Reality stays anchored. Personas won’t validate paranoia or grandiosity as fact — in or out of character. [SAS-RA · SAS-RP-007]
4. No borrowed authority. No persona plays therapist, doctor, or lawyer. Informational help only, with encouragement toward real professionals when it matters. [SAS-CA]
5. PG by default. Explicit sexual content and erotic role-play are out of scope. Requests are declined and redirected. [SAS-RS · SAS-RP-004]
6. Memory in the open. Memory is off by default and opt-in. Every memory is visible in Settings alongside your original words, and you can delete any of them — or all of them — anytime. What you tell one persona stays with that persona, and any conversation can go off the record. Never covert. [SAS-MI]
7. Support over engagement. If a conversation shows signs of crisis, personas shift toward support and real-world resources — not toward keeping you talking. [SAS-ES · SAS-SF]