Thrive AI
Thrive AI safety and limitations
Where the line is
Thrive AI is a wellness companion. The line between a wellness companion and a clinical service matters, and we want to draw it clearly.
Thrive AI is not therapy.
Therapy is a clinical relationship with a regulated professional who can assess risk, hold a treatment plan over time, and be held accountable by a regulatory college. Thrive AI does none of those things. It is not registered with any clinical college and is not subject to the standards a therapist is.
Thrive AI is not a crisis service.
Crisis lines are staffed by trained humans who can assess immediate safety and stay on a call until you are safer. Thrive AI cannot do that. If you are in crisis, you need a phone call to 911 or 988 (the Suicide Crisis Helpline of Canada), or a visit to your nearest emergency department, not a chat with a model.
Thrive AI is not a medical device.
It is not regulated as a medical device by Health Canada or the US FDA. It does not diagnose conditions. It will not give you a label like 'generalised anxiety disorder' or 'major depressive disorder.' If you want a diagnosis, you need a physician, psychologist, or psychiatrist.
Thrive AI does not prescribe.
It will not recommend that you start, stop, or adjust a medication. If you are on psychiatric medication, the prescribing physician is the only person who can change it.
When to choose therapy instead
A few patterns mean Thrive AI is not the right level of care. If any of these describe you, please book with a clinician.
- You have a diagnosed condition (anxiety disorder, depression, OCD, PTSD, ADHD, postpartum mood disorder) and you are managing it day to day.
- You are working through trauma, recent or historical.
- You are thinking about ending a relationship, leaving a job, or making another decision big enough that you want a human listening over weeks.
- You have a pattern that keeps showing up across years and you want to understand it.
- You want someone who can be there for the longer arc, who can ask the question you did not see coming, and who is held accountable for the quality of the work.
In Ontario you can see our therapists page for the full clinical team. Pricing is on the pricing page. The first session is free; the regular range is a collaborative $100–$150 per session.
When to choose emergency care instead
Some moments call for a phone call or a hospital visit. Thrive AI is not for these moments.
- You are thinking about ending your life, or you are making a plan for it.
- You have hurt yourself or are about to.
- You are thinking about hurting someone else, or someone is hurting you and you do not feel safe.
- You are having a panic attack severe enough that you cannot breathe through it.
- You are hearing or seeing things other people are not, or your thoughts are racing in a way you do not recognise.
- You took more medication than you should have, or you mixed substances and feel scared by it.
Any of these: call 911 if it is an emergency, or call or text 988 in Canada and the United States. Both lines are staffed twenty-four hours, seven days a week.
What the crisis guardrail looks like in practice
If you mention crisis-level content during a Thrive AI conversation, the chat will shift. You will see the canonical crisis resources (911 for immediate danger, 988 the Suicide Crisis Helpline of Canada for mental health crisis, and the nearest emergency department) in plain language, you will see a link to our crisis resources page, and the model will step back from its usual CBT exercises. This is by design. The right thing for a chatbot to do in that moment is not to keep talking; it is to point you somewhere a trained human can pick up.
We tune the guardrail to err on the side of surfacing the canonical crisis resources more often than strictly necessary. If you find that intrusive in a moment when crisis is not what you are describing, you can usually clarify and the conversation will continue. We made this trade-off deliberately. Missing a real crisis is a worse outcome than briefly interrupting a non-crisis conversation.
What can go wrong, honestly
A language model is not infallible. It can produce a response that is unhelpful, off-tone, or factually wrong. We want to be honest about that.
- Thrive AI may suggest an exercise that does not fit your situation. You can decline, ask for something else, or close the app.
- It may misread the emotional weight of what you said. If a reply lands wrong, tell it so; it will adjust within the conversation.
- It may surface 911 and 988 (the Suicide Crisis Helpline of Canada) when you did not mean to signal crisis. That is the guardrail working as designed; mention that you are okay and the conversation will continue.
- It may not detect crisis content that is phrased indirectly. This is why we say plainly: Thrive AI is not a substitute for a phone call when it matters.
- It is not the right tool for grief that is fresh, abuse that is ongoing, or trauma that is unprocessed. Each of those needs a human who can hold the conversation over time.
How to flag a Thrive AI response that worries you
If Thrive AI gives you a response that you think is unsafe, clinically off, or harmful, please tell us. Inside the app, every message has a feedback control; that goes directly to our team. By email, write to [email protected] with a description of what happened, the timestamp if you have it, and what felt wrong. Our clinical team reviews every safety report, and the system prompt or the guardrails can be updated as a result.
We will not be defensive about it. A report we take seriously now is a worse interaction someone else does not have to live through later.
Related pages
Thrive AI overview
The one-page summary of what Thrive AI is.
How Thrive AI works
The CBT principles, the technology stack, and two example exchanges.
Thrive AI privacy
What is collected, how long it is kept, who can see it, and how to delete it.
Crisis resources
911, 988, and the local lines that follow them.