Last month, I started teaching my very first college course. Everyone warned me that it was a lot of work, and I absolutely believed them. But, wow, it’s still so much more work than I anticipated! On top of that, I’ve had a couple of other projects that came up in the last month, along with some life things. So I got behind on this newsletter. I wanted to send out something quick for now, and then I’ll try to get back on a bi-weekly schedule later in October.
In honor of the course I’m teaching this semester, which covers AI ethics, today I’ll write about why I encourage caution when using AI as a tool for therapy.
A quick note on terminology:
“AI” is an umbrella term that covers many different types of AI technologies and capabilities. For the most part, when people turn to AI instead of or in addition to therapy, they’re using “large language models,” more commonly know as “LLMs”. ChatGPT, Claude, and Gemini are all examples of LLMs. Unless I specify otherwise, I will be using “LLMs” and “AI” interchangeably in this post, but LLMs are only one type of AI.
As you may have heard, AI systems seem to have encouraged and triggered some users to commit suicide. These systems are designed to be supportive of whatever you say, even if that includes the idea that your life may not be worth living or ending your life. That seems like a really big reason be cautious about using them for therapy.
I know how devastating it can be to lose your life to chronic illness, and I’m here to tell you that regardless of how much healthier or more active you can get given your current condition, there are always ways to improve your life or feel better about the situation you’re in. The resources I provide on my site can help, but there are plenty of other resources as well.
Additionally, AI has triggered delusions and given people incorrect or inaccurate advice. So while it may be helpful for you to talk to the AI, be careful trusting what any AI says to you.
I believe that JournalSpeak is a critical component of recovery, but it is not for the faint of heart. Most likely, you will be writing deep, dark, shadow thoughts that should never be seen by anyone else. I deleted everything immediately after a JS session because I didn’t want to risk that anyone else could see it.
With LLMs, you cannot be certain that your private venting or grief sessions will be kept private. As this article points out, “All of your conversations with ChatGPT are stored on OpenAI's servers.” In general, LLMs are not considered safe and trustworthy tools for sensitive data and information, and even OpenAI acknowledges that it cannot guarantee privacy for free users — it claims it can for paid subscribers, but I would still exercise extreme caution, even with paid services.
If you want to heal, you need to feel safe acknowledging all of your deepest, darkest thoughts and emotions. I, personally, would not feel safe enough on any AI program to acknowledge some of the darkest thoughts I’ve had, and I would not have been able to heal without acknowledging and feeling these things. (Remember, this is about acknowledging and feeling these, and not acting on them or hurting yourself or others.)
Ultimately, all of the tools I recommend for recovery are all about learning how to communicate with your inner self. When you’re in touch with your inner self, then you can make choices and decisions that honor who you are and what you need; you can process your emotions in real time (for the most part); and you can learn how to recognize when your nervous system is getting activated and how to calm yourself down in real time (again, for the most part).
This is a challenging skill to learn, and it takes a lot of practice. You need to be able to tap into the caring, compassionate voice within you to get through challenging life situations, when it may be inappropriate or not possible to turn to an AI for help. You need to learn what questions and suggestions you can present to yourself to help you figure out what’s going on internally and how you can best respond.
Working with a real therapist, reading books about psychology and healing, doing JournalSpeak, practicing somatic meditations — all of these things helped me learn how to communicate with my inner self. It was this broad skillset that enabled me to heal and function again in the real world. Various AI programs may be able to help, but ultimately, this is work that you need to be able to do on your own.
Learning the inner dialogue for healing is tricky and challenging, and some AI programs might help some people access deeper, repressed emotions. I know some people seem to have found AI programs helpful, so I won’t completely suggest ruling them out.
However, I advise exercising extreme caution if you’re going to use any AI programs. As someone who has been in the field of AI ethics for about a decade, I, personally, would not use AI for therapy.
No posts

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.