California Vetoed Its AI Therapy Bill (SB 903). What California Therapists Can Use AI For Now

By The Professions Cloud Editors Updated October 2026 · 8 sources · 6 min read

Prices and facts checked October 28, 2026. We don't accept payment for rankings.

What happened: on September 30, 2026, Governor Gavin Newsom vetoed SB 903, a bill that would have set strict rules on how California therapists and companies use AI in psychotherapy. The same day, he signed AB 1979, a narrower law that stops health facilities and medical offices from letting AI independently do work that requires a license (SB 903 status, AB 1979 status).

Why it matters: California had looked set to join Illinois, Nevada and other states that limit AI in therapy. For now it hasn't. California therapists don't have a therapy-specific AI law, but the bill's author plans to try again, and the existing rules on records and consent still apply.

What SB 903 would have done

SB 903, by Senator Steve Padilla, passed both houses at the end of August, including a 40–0 Senate vote to concur in the Assembly's amendments. Its final text would have (SB 903 bill text):

• limited AI to "administrative" or "supplementary" support. Administrative meant scheduling, billing and logistics messages. Supplementary included progress notes, tracking client progress, and finding referrals, as long as the clinician kept responsibility for clinical decisions.
• required consent before AI recorded or transcribed sessions or did triage or screening. Clients would have been told AI was being used and why, and could say no without losing care. Accepting general terms of use wouldn't have counted as consent.
• banned advertising companion chatbots as therapy.
• required a licensed professional to review and approve before AI could make therapeutic decisions, produce assessments, diagnoses or treatment plans, detect emotions or mental states, or perform triage or screening.
• put responsibility on employers that require clinicians to use AI tools, and protected clinicians from discipline for tool features outside their control.

Why Newsom vetoed it

In his veto message, the Governor said he supported guardrails on AI in healthcare but called the bill "overly broad." He objected that it would (SB 903 status and veto message):

• limit clinicians' use of tools that already help deliver care, including by requiring direct approval of routine screening determinations
• sweep in general-purpose AI through a broad definition of "psychotherapy services"

He asked the author and stakeholders to revisit the issue next year. Padilla has said he will reintroduce it, and the 2027–28 legislative session convenes on December 7, 2026 (Sheppard Mullin via National Law Review).

The bill was backed by professional associations for psychologists, therapists and counselors, and by the National Union of Healthcare Workers. Tech trade group TechNet opposed it, arguing it would put a "clinician bottleneck" in front of intake and screening tools (CalMatters).

What AB 1979 does instead

AB 1979, by Assemblymember Mia Bonta, became Chapter 854 of the Statutes of 2026. California laws generally take effect on January 1, so expect it to apply from January 1, 2027 (AB 1979 text, National Law Review). It:

• applies to health facilities, clinics, physicians' offices and group practice offices, not to every provider
• bans using AI to independently perform any clinical function that legally requires a license, or to direct unlicensed staff in doing one
• requires those settings to make sure licensed providers can still use independent judgment when AI clinical decision support informs care
• doesn't cover documentation and communication tools that don't involve professional judgment, such as reminders and record updates
• treats businesses offering health care chatbots as health care providers under California's medical privacy law, the Confidentiality of Medical Information Act (CMIA)

Does it cover a private therapy practice? It depends on how your practice fits the law's definitions, which come from the Health and Safety Code. A group practice or a clinic setting is more likely to be covered than a solo office. If you're unsure, ask a healthcare attorney.

So what can California therapists use AI for now?

There's no California law that limits therapists to administrative AI the way Illinois and Nevada do. That doesn't mean anything goes:

• Your licensing board's standards still apply. You're responsible for every note, diagnosis and treatment decision, whatever tool helped.
• Privacy law still applies. HIPAA and the CMIA govern client records, including notes an AI drafts. Use tools that sign a business associate agreement (BAA). See our guide to AI note takers and HIPAA.
• AI-written patient messages need a disclaimer in health facilities, clinics and medical offices. Existing California law requires generative AI messages about clinical information to say AI wrote them and explain how to reach a human (SB 903 bill text, legislative digest).
• Companion chatbot rules under SB 243 apply to the companies that run them. See our guide to AI companion chatbot laws.

What California practices should do now

1. Act as if SB 903's consent rule were law. Get written consent before using an AI scribe, explain what it does, and let clients opt out. It's good practice, it's required in several other states, and it's likely to come back. Our AI therapy notes consent form is a free starting point.
2. Keep AI out of clinical decisions. Use AI to draft, and decide yourself. Review every note before signing it.
3. Check your employer's tools. If you work in a clinic or group practice, ask how its AI tools fit AB 1979 before January 1, 2027.
4. Check your clients' states. If you see clients in Illinois, Nevada, Colorado or other states through telehealth, their rules may apply. See our state AI therapy laws guide and 2026 state map.
5. Watch for the new bill after December 7. A narrower version could exempt routine screening but keep the consent and review rules.

How California compares

Sources: Baker Botts, National Law Review. Utah was the first state to legislate on mental health chatbots, in 2025 (Pluribus News).

FAQ

Did California ban AI therapy?

No. The Governor vetoed SB 903, which would have banned advertising companion chatbots as therapy and limited therapists' AI use, on September 30, 2026. Its author plans to reintroduce it in the 2027–28 session.

Can California therapists use AI scribes?

Yes. No California law specifically restricts them, but HIPAA, the CMIA and your board's standards apply. Getting written client consent is still the safest practice, and it would have been required under SB 903.

What does AB 1979 do?

From January 1, 2027, it bars health facilities, clinics, physicians' offices and group practices from using AI to independently perform clinical functions that require a license. It also brings health care chatbot businesses under California's medical privacy law.

Will SB 903 come back?

Its author has said he'll reintroduce it. The Governor asked for a narrower bill that keeps professional judgment in place without limiting useful tools, so expect changes to the screening and review rules.

Changelog: Oct 28, 2026: first published.