Informing clients
Therapists secretly used ChatGPT with their clients
In September 2025 MIT Technology Review spoke to clients who discovered by accident that their therapists were using ChatGPT. In one case, the client's own words had gone into the service without his knowledge.
An article in MIT Technology Review on 2 September 2025 described several cases in which clients found out by accident that their therapist was using artificial intelligence.
During an online session with a 31-year-old client in Los Angeles, the connection was poor, so they switched off video. The therapist then shared his screen by mistake, and the client watched as what he said was typed into ChatGPT and the therapist picked from the suggested answers. The client raised it at the next session; the therapist cried and said he had felt they had hit a wall and had looked for answers elsewhere. The client was still charged for the session.
A 25-year-old client wrote to her therapist about the death of her dog. At the top of the reply, a line of the instruction given to the AI had been left in, promising a more human, heartfelt version. The client felt betrayed, all the more because trust issues were part of why she was in therapy. The therapist apologised: she had never had a pet and had turned to AI to find the right words.
Experts interviewed said general-purpose chatbots such as ChatGPT do not comply with HIPAA, the US health privacy law, and that removing names and addresses is not enough, because sensitive details can often be inferred from the rest. One researcher said many users wrongly believe ChatGPT is HIPAA compliant. A 2023 Cornell study found that AI-assisted messages increase a sense of closeness only as long as the recipient does not know a machine was involved.
What can we learn from it?
In the first case the client's words went to a cloud service without his knowledge; in the second, trust was damaged; honest information given in advance could have prevented both. The lesson is not that professionals should avoid AI, but that they should not use it in secret. In Confiella the recording never leaves the computer, the transcript is made locally, and only reviewed, pseudonymised text goes to the cloud, but that does not replace informing the client. Good practice is to say at the start whether you use AI tools, for what, and what leaves the consulting room.
Sources
- MIT Technology Review: Therapists are secretly using ChatGPT. Clients are triggered.
- Futurism: Patients Furious at Therapists Secretly Using AI
This article is based on the public sources listed. If you find a mistake, write to us: [email protected]