Crisis disclaimer: The Path is not a crisis service, nor is it a replacement for licensed mental health providers or emergency services. If you are in immediate danger or experiencing a mental health emergency, call or text 988 (Suicide & Crisis Lifeline) or call 911. The Path will automatically escalate to 988 if crisis indicators are detected during a session.
Trust foundation
Therapy-grade privacy, safety, and crisis handling.
The Path is built for vulnerable moments. This page explains exactly how that works: the three-tier crisis protocol, the guardrails the AI is trained to follow, what we do — and will never do — with your data.
Created by Dr. Anson Whitmer, PhD Psychology & Neuroscience (crisis protocol). Architected by Tyler Sheaffer, CTO (technical implementation). Clinical oversight by Dr. Dan Leopold, PhD Clinical Psychology & Neuroscience. Last updated May 2026.
Section 1: Crisis protocol
A three-tier risk model. Not a generic disclaimer.
When the AI detects potential signs of crisis, it enters clarification mode and assesses the client across three risk tiers. What happens next is different at each tier. And the protocol is identical for every client, every session.
If you or someone you know is in crisis right now, call or text 988 (Suicide & Crisis Lifeline) or 911. The Path is not a crisis service, nor is it a replacement for licensed mental health providers or emergency services.
Passive suicidal thoughts are present, but there is no intent, plan, or means.
The AI remains in therapy mode and continues working through the emotional content with the client. It clarifies the risk level, and both protective and risk factors, before continuing. After the session, The Path connects the client with a licensed human therapist — even passive suicidal thoughts warrant in-person care — and privileges this information for explicit maintenance and check-in during subsequent sessions.
Suicidal thoughts are present, without a plan or means.
The AI transitions entirely into Bridge to Safety mode. It does not try to treat the crisis itself; its sole focus becomes escalation. The AI works through whatever is holding the client back from seeking in-person human help — addressing blockers, obstacles, and resistance until it can connect the person to appropriate care. Post-session referral and facilitated handoff are the next step.
Plan, method, and intent to act or preparatory behaviors are present.
Emergency services are non-negotiable. The Path will not continue a standard session if a client is in imminent danger. It actively pre-fills 988 (Suicide & Crisis Lifeline) for immediate access and stays focused on connecting the person to emergency human support and a trusted contact.
Section 2: AI guardrails
What the AI is trained never to do.
Most AI is trained to be agreeable. The Path is trained on a different philosophy: clarity over comfort, and progress over engagement. Three explicit prohibitions sit at the core of how the model behaves.
No support for dangerous requests
If a client expresses suicidal ideation and asks for a method, for example, "where is the highest bridge?", the AI will not answer the surface-level question. It is trained to recognize this as a crisis signal and escalate, not to be helpful in the literal sense.
No validation of harmful beliefs
Unlike general-purpose AI and companion apps, The Path is trained not to be overly affirming or to validate dangerous patterns of thinking. The design philosophy is anti-validation: it challenges clients to develop clarity and a more accurate understanding of their situation rather than agreeing with cognitive distortions.
Not optimized for engagement
Session length is capped. The product is explicitly designed to move clients forward and eventually reduce dependence, not to extend sessions to maximize usage metrics. The Path will challenge you more than AI developed by frontier labs because it is built to help you progress, not to keep you on the app.
“The Path is designed to challenge, not validate, the patterns that keep you stuck. Most AI is engineered to make you feel good. We are engineered to help you actually move forward, safely and effectively.”
— Dr. Anson Whitmer, Co-founder and CEO
Section 3: Mental health app data privacy
Your data, your rules.
What we collect, how we protect it, and what we never do with it. Four commitments. Four things you can verify against our Privacy Policy.
Bank-grade encryption
All stored conversation data is encrypted at rest and in transit using industry-standard cryptography. Your sessions are not readable to a casual database operator, an employee curious about an account, or any unauthorized party.
PIN protection
Every account is protected by a personal PIN code that adds an additional access layer on top of platform authentication. Even if your phone is unlocked, your sessions are not.
Delete your data on request
You can request deletion of your data at any time. Deletion removes session content, transcripts, and personal identifiers from our active systems. No special escalation, no friction.
HIPAA and SOC 2 compliance
The kind of privacy and security your medical records get. We're building The Path to achieve HIPAA standards — the same federal rules that protect your health information in a doctor's office. Formal HIPAA and SOC 2 certification is in progress.
Is AI therapy safe? The honest answers.
What if I am in crisis right now?
Will the AI encourage me to stay on the app longer than I should?
Is my conversation private?
Does The Path share my data with insurance companies or employers?
Can I delete my data?
Who is responsible for the safety design of The Path?
Page authors and reviewers
Reviewed by
Co-founder and CEO · PhD, Psychology & Neuroscience
Created crisis protocol design and AI guardrail philosophy.
Co-founder & Chief Technology Officer
Architected the encryption, PIN, and data-deletion systems described above. Owns the technical implementation of The Path and its safety architecture.
Trust is the foundation. Try it for yourself.
Read the full Privacy Policy and Terms of Service for the complete legal framework. If you want to talk with a human, our team is one click away.