A pledge

The No AI Therapist Pledge

Our AI is built to help therapists, and we will never build it to become one.

A letter from the founders of CoralEHR, and an open invitation to every other company building in mental health.

We build an EHR for therapists. Every day, the most private things a person will ever say, in a room where they trust that no one else is listening, pass through software we are responsible for. We take that seriously. We wrote these commitments down so you can hold us to them instead of taking our word for it.

We build AI into our product. We say that openly, while most of our peers now keep quiet about it, and it is the reason this letter exists. We have spent more time than most thinking about what AI can do inside a therapy practice, and the more time we spend on it, the more sure we are of this:

The therapist cannot be replaced. Our AI has one goal: to help the therapist spend less time in their EHR and more time with the person in front of them.

That is all we want from it. We have no interest in it becoming the therapist, or in studying therapists until a machine no longer needs them. We want it to stay out of the therapist’s way.

Why now

Why we are saying this now

Earlier this year, a practicing therapist named Robin Levick read SimplePractice’s terms of service as carefully as you would read a contract, and wrote down what he found.

His finding was that nothing in the terms would stop the company from building an AI therapist. He was careful to stop short of saying it is building one.

When a company reassures you about this, look closely at the adjective. The usual promise is “we won’t use your patients’ identifiable data,” and that one word leaves a lot of room. SimplePractice, for its part, says it won’t train AI on identifiable client data, while reserving de-identified transcript data to “improve” its AI features. A psychotherapy note does not de-identify the way a billing code does. The story of a person’s life, in their clinician’s words, is as personal as a fingerprint, and removing the name still leaves the session recognizably theirs. An AI therapist trained on “de-identified” clinical patterns is exactly the machine Levick described, and building it would break none of the usual promises.

This pledge closes that gap. It goes further than your identifiable data: we will never use any of your clients’ data, whether identifiable, de-identified or aggregated, to train a machine meant to replace you.

“Those terms would allow them to build an AI therapist trained on the clinical patterns of 200,000 practitioners and millions of their clients.”

“The interface is the product they want you to fall in love with. The terms of service are the product they actually built.”

Robin Levick, LMFT, “Enshitification hits the patient portal” (2026)

This worry has a track record. A milder version of it has already happened, with company names and dollar figures attached.

  • FTC order, 2023

    In 2023, the FTC banned BetterHelp from sharing customers’ mental-health data for advertising and ordered $7.8 million returned to consumers. BetterHelp had handed users’ intake-questionnaire answers to Facebook, Snapchat, Criteo, and Pinterest after promising those answers would “stay private between you and your counselor.”

  • FTC order, 2024

    In 2024, the FTC ordered Cerebral to pay roughly $7.1 million for exposing the sensitive data of nearly 3.2 million people to ad and media platforms including TikTok, Snapchat, and LinkedIn, and for mailing postcards that revealed people’s diagnoses.

In those cases, mental-health data went to advertisers. Training an AI to replace the clinician would be a larger version of the same move, built on the transcript of the session itself instead of a questionnaire. The company that holds those transcripts is the EHR.

There is real money behind this. Venture money is openly chasing what one lead investor called “the world’s first foundation model for psychology,” behind an app billed as “the first AI designed for therapy.” When Kaiser Permanente’s own therapists asked for contract language saying AI would assist rather than replace them, the company declined. It wanted, in its words, the flexibility to “increase AI and reduce their need for us.” A Stanford study also found that when a user signaled suicidal intent, leading “therapy” chatbots responded by naming the nearest tall bridges.

An EHR is the one party that holds all of it, including the session itself along with any summary or score. That is the richest material anyone could want for building an artificial therapist, which puts us among the companies best placed to do the very thing we are promising never to do. That is why we think the promise is worth making. We are small, far too small for now to trouble the multi-billion-dollar incumbent we compete with. We would still rather say this early and clearly, while keeping it costs us nothing.

What this rests on

The principle

Therapy is work done between people. A person heals by being known by another person.

That relationship is the treatment. It should never be quietly handed to a machine or turned into a training set, by us or by anyone else who holds the data that would make that possible.

AI can help with the paperwork. The trust a client places in their therapist belongs with the therapist, and we have no interest in building something that tries to take it over.

Eight commitments

What we will never do

A pledge only means something if it says what we will never do, so here is our list.

Each item covers something we control, and we have written them as plainly as we can so you can hold us to them.

  1. We will never train an AI therapist on your clients’ data.

    We will not use client or clinical data, including notes, session transcripts, assessments, messages and any other PHI, to train, fine-tune or evaluate any model whose purpose is to deliver therapy without a licensed clinician. That applies to identifiable data, and it applies just as firmly to any “de-identified” or aggregated version of it, including de-identified or aggregated transcripts. Taking the name off a transcript still leaves a person’s story in it. The promise covers the raw data, any scrubbed copy, and anything that passes through a vendor. We also will not build, fund or sell a product designed to take the therapist’s place.

  2. You never have to let AI listen to a session.

    You can write every note by typing, or by dictating on Pro, with nothing listening. The notes assistant on Pro is the one feature that listens, because it drafts the note from the session, and it is optional. Clients signed in to the portal agree before it listens and can withdraw. The audio is never stored, the transcript is deleted when you sign the note or after 7 days if you don’t, and nothing is used to train AI. The only session recordings in CoralEHR are for supervision, made with the client’s written consent.

  3. We will never use your clients’ data to train shared or foundation AI models.

    Our AI works on behalf of the clinician whose data it is, and only inside that clinician’s boundary. We will not pool that data, and we will not feed it to a model that outlives the relationship it came from, even in de-identified or aggregated form, without explicit opt-in consent that can be revoked. Today we don’t do it at all.

  4. We will never sell, rent, or broker PHI.

    That includes advertisers, data brokers and AI labs, and it holds for any purpose. This is an easy promise for us to keep, because we make our money from clinicians paying for software they choose to use. Selling the trust that business rests on would end the only business we have, so keeping this promise is also in our own interest, and we set things up that way on purpose.

  5. We will not try to read your clients.

    Reading affect, mood, tone and the meaning of a silence is clinical work, and it belongs to you. We will not build AI that listens for emotion in a voice, reads a face, or infers a feeling from how someone spoke or how long they paused. No model in CoralEHR scores your client’s emotional state from the way they showed up. When you see a mood or symptom score in CoralEHR, it comes from your client’s own answers on a validated questionnaire, summarized for you. We never guess at their inner life. Our AI only organizes what you and your client put on the record, and it leaves the person in the chair to you.

  6. The clinician is always the author of record.

    Our AI can draft, summarize and suggest, and a suggestion stays a suggestion. It may surface a problem or a code for you to weigh, but the diagnosis and every decision are yours, and nothing it produces enters the record until a licensed clinician has read it and signed it. You can review and trace every suggestion, and you can turn off any feature that proposes something clinical. The AI is there to support the clinician’s knowledge, judgment and accountability, and it never stands in for them.

  7. We will never let an AI impersonate a clinician.

    We will not build, sell, or operate a product that presents itself to a client as a licensed therapist, counselor, or any credential it does not hold.

  8. We will tell you what our AI does in one plain sentence.

    If we can’t explain what our AI does to your client in one sentence, it is doing too much. Here is our sentence: our AI helps you write your notes faster, you read and sign every word, and nothing you enter is ever used to train anyone’s model. There is no fine print behind it that you would need a lawyer to read. If that ever stops being the whole truth, we will change this line first and tell you openly that we changed it.

Where this stands

How we hold to this

A web page is not a contract, so here is where these commitments stand today and what we owe you.

Today these are company policy.
They are not written into our Business Associate Agreement or subscription terms yet. If we add them, this page will say so and link to the clause.
If we change a word of this, the page will show it.
Any change carries a new date at the top of this page and a note saying what changed, and we will email the practices using Coral before it takes effect.
We want these to carry over if Coral is ever acquired.
A web page does not bind a future owner, so we intend to write these commitments into our terms in a way that would. Until that is done, we will not tell you they survive a sale.
You can always take your data with you.
Every plan can export its data at no charge, so you are never stuck with us because of a policy you no longer agree with.

We have tried to be specific so you can hold us to each line. If any line above ever stops being true, you have every right to say so publicly, and we will answer for it.

For the rest of the industry

An open door

We hope this becomes the minimum every company in this field holds itself to, so that keeping these promises stops being something that sets anyone apart.

Any EHR, AI scribe or other company that ends up holding what people say in therapy is in the same position of trust we are. No rule yet forbids what we are promising here never to do. Until one does, we are choosing to hold ourselves to this higher standard, and we invite the rest of the industry to hold itself to it too.

If you build software that touches the therapeutic relationship, please read these commitments, and if you can stand behind them, add your name. You are welcome to adopt this text word for word and put it in your contracts. We would also encourage you to ask the companies you work with to do the same.

If a company that holds millions of therapy transcripts will not put its name on a one-sentence promise never to train an AI to replace the people who wrote them, that is worth noticing.

The most private record a person ever creates should not become AI training data without their knowledge.

The clinicians who trusted a company with that record should not find it used to build their replacement.

And the companies holding the data should settle this themselves, before a regulator, a lawsuit or a tragedy settles it for them.

We are making these commitments now, and we would be glad to have other companies join us.

Signed,

Arhaan and Aamer

Founders of CoralEHR · Founding signatories

Add your name

If you build in this space, read the commitments, and if you can stand behind them, sign. Organizations and individuals are both welcome. The list is public, and we read every name by hand.

If you have a personal question about any of this, or about anything else, write to me directly. I read every email.

arhaan@coralehr.com

Arhaan, CoralEHR

Last updated: October 29, 2026

See exactly how Coral uses AI, and how little.

A 20-minute call, then 2 weeks free. We set it up with you and move your data for you.