Mental Health AI Compliance California (2026): AB 489 Rules
Mental health AI faces the strictest scrutiny under California law due to the vulnerability of patients and potential for harm.
Check Your Mental Health AI Compliance
Free 2-minute assessment with personalized action plan
AI Applications in Mental Health Providers That Require Compliance
- 1AI chatbots providing therapeutic support or advice
- 2Mood tracking apps offering AI-generated insights
- 3Virtual companions between therapy sessions
- 4Crisis detection and intervention systems
Key Compliance Requirements for Mental Health
- Explicit disclosure: "I am an AI. I am not a therapist."
- Cannot simulate professional therapeutic relationship
- Immediate human escalation for crisis/suicidal ideation
- No use of titles like "Counselor" or "Therapist"
💡 Mental Health Compliance Tip
AB 489 specifically calls out mental health chatbots. Non-compliance risks both regulatory action and patient harm lawsuits.
Running a patient-facing chatbot?
If your mental health providers use a generative-AI chatbot to communicate with patients, AB 3030 has specific disclosure and human-review rules.
Is your medical chatbot legal under California AB 3030?California AI Regulations Affecting Mental Health Providers
Mental health AI is treated as the highest-risk category under California law. AB 489 mandates explicit, repeated disclosure that the AI is not a therapist — and the legislative history specifically references companion-chatbot harm cases. A small "AI" badge is not sufficient; the BBS has indicated it expects opening text along the lines of "I am an AI. I am not a licensed therapist or a human."
SB 243 (the "companion chatbot" law) adds standalone requirements for any AI designed to form ongoing relationships with users — including disclosure at session start, crisis-detection escalation paths, and limits on engagement-driving design patterns. Most "AI therapy" apps fall under both AB 489 and SB 243 simultaneously.
For minors, AB 56 (the social-media black-box warning) and parallel age-assurance rules add disclosure obligations. Crisis detection (suicidal ideation, self-harm) must trigger an immediate escalation path to a licensed human — automated handoffs to a "crisis bot" do not satisfy the standard.
Is Your Mental Health AI Compliant?
Take our free compliance assessment to identify gaps and get a personalized action plan for your mental health practice.
Check My Compliance Now