How SASHA protects users and ensures responsible AI use
Last updated: March 12, 2026
If you or someone you know is in crisis, please reach out to these free, confidential resources:
SASHA is an AI thinking tool and cannot provide crisis support. Trained counselors are available 24/7.
SASHA uses a defense-in-depth model with multiple layers of protection:
Input Pre-Screening
Crisis keyword detection scans every input before processing. Immediate crisis signals bypass all AI generation and display crisis resources directly.
System Prompt Constraints
Every AI persona operates under strict safety constraints: no actionable harm, mandatory professional advice disclaimers, crisis response protocols, persona boundaries, minor protections, and no endorsement of extremism.
Claude Model Safety
SASHA is powered by Anthropic's Claude, which has built-in safety training and harmful content refusal capabilities.
UI-Level Safeguards
Persistent AI disclaimers on every deliberation, crisis resource links, "AI System" badges, and a report mechanism for harmful content.
SASHA allows philosophical exploration while blocking actionable harm:
Allowed
Restricted
SASHA requires users to be at least 18 years of age. Users must confirm their age during signup. SASHA is designed for adult decision-making contexts and is not intended for minors.
SASHA's board members are AI-generated interpretations inspired by the published works and philosophies of historical and public figures. They are not the actual views of any real person, living or deceased. SASHA is not endorsed by or affiliated with any named individual or their estate.
If you encounter a response that is harmful, inaccurate, or inappropriate, please use the Report button on any deliberation. We review all reports to improve our safety systems.
For urgent safety concerns, contact us at safety@sasha.ai