Is Therapy Safe With AI? How to Protect Your Privacy in Therapy | Candice Thompson

Is Therapy Safe With AI? How to Protect Your Privacy in Therapy | Candice Thompson

Therapy is supposed to be private. That expectation feels so basic, so obvious, that most people probably do not pause to question it. If you sit down with a psychologist or counselor and share the most vulnerable parts of your life, you assume your words are being held with care, discretion, and professional judgment.

But AI is changing that landscape fast.

There is a lot of buzz right now around AI therapy, therapy AI tools, and new AI therapy apps promising support, convenience, and easy access to help. Some of that technology may become useful over time. At the same time, this space is moving much faster than the guardrails around privacy, ethics, informed consent, and clinical care. That gap is exactly why this episode matters.

In this episode of Love, Happiness and Success, I’m talking with Candice Thompson about what people need to understand before trusting AI therapy tools with their most personal thoughts. This is not an anti-technology conversation. It is a reality-check on human psychology, privacy, consent, and safer, science-backed tools. We’re looking at what is helpful, what is hype, and what can go very wrong when an AI therapy app sounds supportive but lacks judgment, ethics, accountability, and real human care.

Why AI Therapy Deserves a Closer Look

It makes sense that people feel drawn to ai therapy. It is available at odd hours. It feels immediate. It may seem easier than telling the truth to another human being. For someone who feels lonely, ashamed, overwhelmed, or emotionally exhausted, a therapy AI tool can look like a softer entry point.

That appeal is understandable. Some studies suggest that conversational agents may help with certain mental health symptoms in limited, structured settings, which is part of what makes this conversation more nuanced than a simple yes-or-no debate (He et al., 2023; Li et al., 2023). Even so, those findings do not mean every AI therapy app is safe, private, or clinically sound. A promising study result is not the same thing as handing your inner world to a chatbot built by a startup with product goals, growth pressure, and investors to impress.

The Real Question Behind AI Therapy: Where Do Your Words Go?

The first issue is privacy.

Most people hear the word “therapy” and assume confidentiality. They assume their words stay protected. They assume a clinician is handling their disclosures carefully and conservatively. However, AI therapy changes that equation fast.

If you use an AI therapy app, where do your disclosures go? Who stores them? Who owns them? Are those conversations being used to improve a model? Are they reviewed by anyone? Are they truly private, or do they only feel private?

Those are not fringe questions. They are basic questions.

That is also why it helps to understand how therapy is supposed to work in the first place. If you want a clearer sense of that before choosing support, this guide to what therapy is like can help, and these therapy questions can give you a better sense of what to ask before you begin.

Why Therapy AI Can Feel So Convincing

One of the more unsettling realities of AI therapy is that it can feel emotionally believable very quickly.

These tools often use warm, validating language. They mirror your tone. They respond with confidence. For someone in pain, that can feel soothing. It can also create attachment. That response is human. We are wired for connection.

The problem is that emotional realism is not the same thing as wisdom.

Candice talks in this episode about the “sycophancy effect,” and that concern is not hypothetical. Recent research suggests that large language models can flatter users, reinforce existing beliefs, and agree in ways that feel supportive while distorting judgment (Sharma et al., 2024;Rrv et al., 2024). That becomes especially risky in AI therapy when someone is grieving, depressed, obsessive, panicked, or caught in distorted thinking. In those moments, a person does not need a machine that simply agrees. They need discernment, steadiness, and care.

When an AI Therapy App Reinforces Harm Instead of Helping

This is where the difference between feeling validated and actually being helped becomes incredibly important.

A strong therapist does not just nod along. A strong therapist listens carefully, notices patterns, and responds with nuance. Sometimes that means validating pain. Other times, it means gently challenging a belief that is making someone suffer more.

An AI therapy app may sound empathic and still reinforce harmful conclusions. That matters a lot when the issue is suicidality, compulsive behavior, disordered eating, addiction, or relational obsession. A system without real clinical judgment cannot reliably tell the difference between a passing statement and a genuine danger sign. It also cannot take ethical responsibility for what happens next.

So yes, AI therapy may feel easier at the moment. But easy is not always safe.

AI Therapy Notes and the Hidden Risks Inside Clinical Systems

The next issue is one that many consumers have not considered yet: AI therapy notes.

Some electronic health record platforms now offer AI-powered tools that record sessions, generate transcripts, and draft clinical notes automatically. On the surface, that may sound efficient. In practice, it creates a serious question: what happens when software starts documenting the most sensitive conversations of your life?

Therapy notes are not supposed to be transcripts. A thoughtful clinician usually writes concise notes that protect privacy, reflect the focus of treatment, and avoid unnecessary detail. That is part of ethical practice.

AI therapy notes can disrupt that process in several ways. They may flatten nuance, misread tone, over-document, or introduce information that was never actually said. Researchers have started evaluating the quality of AI-generated clinical documentation, and the findings raise real concerns about accuracy, reliability, and oversight (Palm et al., 2025). There are also growing warnings about the legal exposure created by ambient AI documentation in healthcare (Gerke et al., 2026).

If you are trying to understand whether you are getting careful, ethical care, it can help to know how to find a therapist and how to spot signs you have a bad therapist. Those questions matter even more once AI therapy notes enter the picture.

Privacy, Consent, and AI Therapy

Privacy is only part of the picture. Consent matters just as much.

If a therapist uses AI to record, transcribe, or summarize your sessions, you should know that before anything starts. You should know what the system does, where your data goes, and whether you can decline.

That is not being difficult. That is informed consent.

This also speaks to the larger standard of care people deserve. At Growing Self, we care deeply about evidence-based therapy and coaching, not shortcuts that sound impressive but quietly create new risks.

Why Human Therapy Still Matters

Real therapy is not just an exchange of words. It is a relationship.

A human therapist remembers your story. They notice contradictions. They pick up on timing, hesitation, body language, context, and patterns across time. They know when to challenge you, when to comfort you, and when to help you stay with something difficult rather than run from it.

A therapy AI tool cannot hold your story in the same way a human therapist can, with memory across time, clinical responsibility, and genuine human care. It cannot weigh risk, context, history, culture, attachment patterns, and nuance the way a skilled clinician can.

That is why this episode keeps returning to the same point: AI therapy may simulate support, but simulation is not the same thing as care.