Is AI Therapy Private? How to Check Before You Overshare
You are about to tell an app things you have not told people who love you. Before you do, spend five minutes on this page, because the mental health app industry has already burned this trust once, at scale, and the receipts are public. Here is what actually happened, the five questions that separate careful apps from careless ones, and how sokoon answers each, so you can hold every app, including ours, to the same standard.
This industry has a record, and you should know it
In 2023 the US Federal Trade Commission reached a $7.8 million settlement with BetterHelp over sharing the health information of users, including the fact that they were seeking therapy, with Facebook and Snapchat for advertising. The same year, telehealth company Cerebral disclosed it had exposed the health data of over three million patients to advertisers through tracking pixels. These were not hackers. These were the products working as built.
Mozilla's research team reviewed dozens of mental health apps and slapped privacy warning labels on the large majority of them. The pattern is consistent: the industry that asks for your most sensitive data has, on average, treated it most carelessly. That is the context every AI mental health app operates in, and it is why "trust us" is not an answer. Only specifics are.
The five questions that actually matter
- Is it encrypted? In transit and at rest. If the word encryption does not appear in the privacy policy, assume the answer is no.
- Do your conversations train AI models? This is the new one people forget. Many AI products use your chats as training data by default. For a mental health tool, that means your worst nights could become someone's model weights. Look for an explicit promise that they do not, with no "to improve our services" wiggle room.
- Is data sold or shared with advertisers? The BetterHelp question. Look for tracking pixels, ad partners, and the phrase "may share with third parties."
- Can you delete everything, easily? Not "email support to request deletion." A button.
- What happens in a crisis? Not strictly privacy, but the same honesty muscle. An app that is vague about crisis behavior is usually vague elsewhere.
How sokoon answers them
Conversations are encrypted, never sold, never shared with advertisers, and never used to train AI models, ours or anyone else's. Deletion is yours: conversations and the whole account, whenever you want. Signup asks for an email and nothing else, no real name, and the free plan never asks for a card. In a crisis, sokoon stops and points to human help, like 988 in the US. The full detail lives in our privacy policy, which we tried to write like we expected you to actually read it.
We put those answers in writing on public pages on purpose. When a company states a specific promise in public, walking it back has a cost. Vague language is cheap. Specifics are collateral.
A note on general chatbots
If your mental health conversations currently happen in a general purpose assistant, the privacy question changes shape. Consumer AI assistants may use your conversations for training unless you opt out, and their policies are moving targets. None of that is scandal, it is just a product built for a different job. The full comparison is in sokoon vs ChatGPT, and the wider picture in using ChatGPT for mental health.
The bottom line
Anonymity and privacy are different layers: one is whether the app knows your name, the other is where your words go. We covered the first in how anonymous an AI conversation can actually get. For the second, use the five questions above on any app you are considering. An app that answers all five plainly has done something rare in this industry. An app that answers none of them has answered you anyway.
Say it somewhere built to keep it.
Encrypted, never sold, never training data, deletable whenever you want. One free session every day, starting tonight.