Calm AI Therapy Editorial · 7 min read

Who Owns Your Mental Health Data? What Every AI Therapy App Does With Your Pain

Most AI therapy products treat your conversations as training data. We pull back the curtain on the privacy landscape and explain Calm AI Therapy's commitments — in plain language.

Published March 28, 2026

When you sign up for an AI therapy product, the most important question is rarely asked: where does the thing I just said go? The answer, on most platforms, is "into the model." The conversations you have are stored, indexed, and — depending on the product's terms — used to train the next version of the AI. Your worst day becomes a training input.

This is not a minor concern. The thing that makes therapy work is the assumption that what you say does not leave the room. Once you puncture that assumption, the conversation changes. People hold back. They self-edit. They never quite get to the thing that mattered. The product becomes useless to its own users without the users noticing why.

The legal landscape is murky. AI therapy is not regulated under HIPAA in most cases, because the products themselves are not classified as medical providers. Some hide behind ToS clauses that allow data resale, training use, and third-party sharing. Others claim privacy in marketing copy while their actual practices are different. There is currently no single, enforceable standard.

Calm AI Therapy was built around a different premise. We do not train on your conversations. We do not sell your data. We do not share it with third parties. These are not aspirational claims — they are architectural commitments, baked into the system. Sessions are encrypted. Memory is per-user and isolated. You can export everything you've ever said in one click. You can delete everything in one click. None of that is opt-in.

We are aligned with HIPAA's data principles even where we are not legally required to be, because the people who use our product deserve at least that. We will publish our subprocessor list. We will publish our incident response policy. We do not bury data terms in 40 pages of ToS. The terms that matter are stated on a single page, in plain English.

There is a broader question here about who builds these products. The teams behind some AI therapy apps are venture-funded with growth at any cost. The growth model often rewards engagement metrics that are at odds with user wellbeing — and it rewards data accumulation. Calm AI Therapy does not carry those incentives. There is no investor here expecting a return on your attention. We are not trying to maximise daily-active-users. We are trying to be useful.

If you've ever paused before typing something into an AI tool because you weren't sure what would happen to it — that pause is wisdom. Most of the time, the right answer is: don't type it. Calm AI Therapy exists for the cases where the answer should be: it's safe here. We have committed to keeping that promise. The page that explains exactly how is one click from this article.