Therapists secretly using ChatGPT: patients’ private data could be at risk

Therapists are turning to ChatGPT in secret, raising alarms over patient trust, privacy, and whether AI belongs in therapy at all.

I used to be a regular therapy-goer, and I would have been mortified if my shrink had been using AI during our Zoom meets.

Luckily, my therapist was all about riding off the subject matter we’d both produced in the moment, though I’m sure some nifty specialists would still be able to shift to an LLM, sub rosa.

For me, my therapy sessions were built on imperfections, and no, I don’t mean me being damaged goods, but the ebb and flow of a curative rivulet.

As the MIT Technology Review recently explored, patient trust collapses when your therapist might be using contrived responses when using AI – they’d better be careful when sharing the screen!

Cybernews spoke to a couples’ and individual therapist, Thomas Westenholz, to help gauge the ethics after all.

Wetenholz believes it should be a case of transparency on the therapist's part:

“Therapy is about honesty and openness. If a client discovers their therapist has been relying on ChatGPT behind the scenes without disclosure, it risks breaking that bond of safety.”

My therapist would keep me in the loop with her back-to-back schedule. In fact, I wouldn’t be surprised if she had somewhere between six to eight clients in a working day, from the range of slots she offered me.

If a counsellor is feeling burned out or the turnaround time for a particular problem is particularly time-sensitive, it’s not difficult to see why they would take shortcuts, especially when something is brought up in the moment.

However, solutions are often found together, with the client – and if some scripted notes were being read off to me, I’d feel let down.

Sure, AI could occasionally come in to prepare some guidance before a session, but like a teacher, it should remain mainly for admin tasks.

“AI may have a supportive role, helping therapists with admin, generating psychoeducational resources, or providing prompts for reflection. But it should never replace the human presence that therapy depends on,” shared Westenholz.

A study this year pointed out that AI responses may be better suited to the client's needs, but with the caveat of them not knowing that it was synthetic.

However, the element of trust is the binding thread that weaves through psychotherapy meetings, and a patient later may feel deceived if they realize their issues were being input into an LLM.

A client may in fact feel even more vulnerable, especially given the fact that high confidentiality is often a key agreement in the beginning between the client and professional.

“Transparency is non-negotiable. In EFT, we understand that healing occurs in relationships where individuals feel secure, and that security is dependent on truthfulness,” explains Westenholz.

“If AI is part of the process, clients deserve to know and to consent.”

Unlock more exclusive Cybernews content on YouTube.