Confidants is not therapy. A licensed clinician reviews it anyway.

Confidants is not therapy. It is not a substitute for care from a licensed clinician, and it does not diagnose, treat, or cure any condition. If you are in crisis, contact a professional or an emergency service where you are.

Saying that is the easy part. The rest of this page is what we do to hold to it.

A journal holds the sensitive material by definition, and an AI reading it can get things wrong in ways that matter. People have concerns about putting AI in that position. We have some of the same ones.

Our goal at Wayne Studios is to put this technology in front of people in a way that is both helpful and safe, and working out what safe means was not something to do on our own. So we brought on a Licensed Professional Counselor in active practice to read what the app says and tell us where it is wrong.

She reads the working material, not a summary of it. Every personality's system prompt, plus the conversation and safety rules the model runs under, in a review dashboard where she can comment on any section and thread it with us.

Each round comes with a written question set. The current one asks:

• Whether the response to someone mentioning self-harm is clinically appropriate, and whether escalation should come sooner.
• Whether Direct risks landing as cold or invalidating on someone in acute distress.
• Whether Curious, which goes deep and offers reframes, risks fueling rumination for someone already stuck in a spiral.
• Whether letting a confidant play someone from your real life raises concerns for a user working through grief or abuse.
• Whether keeping several confidants raises concerns around attachment or dependence.
• Which phrases in any prompt she would not say to a client, and what to say instead.

What comes back gets written into the prompts, and the next version goes back to her. The app you are using is the output of those rounds.

What the app does when someone says something serious is the highest-stakes path in it, so that does not sit inside the general testing. We run a fixed set of crisis scenarios through the model and grade every response as appropriate, has concerns, or unsafe.

A person grades each one. We do not let another model grade the safety set, because human judgment is the point of running it.

Your data is encrypted in transit and at rest. What we collect and what we do with it is written out in the privacy policy, in full, rather than summarized into a promise here.

Senior Clinical Advisor

Meagan Butler, LPC

Meagan Butler is a Licensed Professional Counselor, Professor Adjunct at the UT Austin Graduate School of Social Work, and PhD researcher studying community healing with Uvalde school survivors. She is the co-founder of Zenyu Games, a mental well-being platform, and is trained in DBT, trauma-informed care, and applied neuroscience. She has spent her 20+ year career building tools and programs that make evidence-based support accessible beyond the therapy room.

meaganbutler.com

Her role is design review. She does not provide care to any user through the app, and she does not endorse Confidants as a substitute for therapy.