Skip to content
Thu 19 Nov 2026 Toronto Sign in
Policy

Newsom Wouldn’t Ban AI Therapists. He Signed the Clinic Version.

On his last signing deadline as governor, Gavin Newsom vetoed SB 903's limits on AI in psychotherapy while signing AB 1979, which keeps AI out of licensed clinical functions.

An empty therapist's office with two chairs facing each other
Illustration: prompt/power

State Sen. Steve Padilla had a line ready for his AI therapy bill, and it was a good one: “AI can be a tool in the hands of licensed professionals, but it cannot be the professional itself.” His office put it in an Aug. 31 release as SB 903 cleared the California legislature. On Sept. 30, the last day Gavin Newsom could act on this year’s bills, the governor vetoed it.

Updated Oct. 19, 2026: AB 1979 and SB 503 take effect Jan. 1, 2027, with AB 1979 enforced by California’s professional licensing boards, according to the California Health Care Foundation’s Oct. 8 review. As of Oct. 19 we found no lawsuit or regulator guidance on either law, and no new bill on consumer AI therapy apps.

Newsom Wouldn't Ban AI Therapists. He Signed the Clinic Version.
Image: Office of the Governor of California

The same afternoon he signed a different bill that says something very close to Padilla’s sentence, as long as the AI is inside a clinic. That split is the clearest way to read Newsom’s final batch of AI laws before he leaves office.

The therapy bill that died

SB 903 would have limited AI to “administrative or supplementary support in psychotherapy services,” according to CalMatters’ Digital Democracy summary. It barred AI from triaging patients or making therapeutic decisions without review by a licensed professional, required patient consent before AI was used in their care, and banned advertising psychotherapy services “provided through the use of companion chatbots.”

We could not extract text from the veto message PDF the governor’s office posted, so we are not quoting Newsom directly. The California Health Care Foundation’s summary says he found the bill’s definitions and scope broad enough to harm legitimate clinical tools never meant to replace psychotherapy. Padilla’s office described it in an Oct. 1 statement: “The Governor’s veto message claims the protections for patients and clinicians in this bill go too far, but Governor Newsom knows all too well the dangers chatbots pose to vulnerable Californians, and despite that knowledge, has allowed unlicensed algorithms to act as therapists.” That is the bill author’s account of the veto, and it should be read as one.

A second health measure went down too. AB 2575, Assemblymember Liz Ortega’s bill on AI in health care services, was vetoed, per the governor’s Sept. 30 legislative update and a tally by the privacy law firm Stauss.

What he signed for hospitals

AB 1979, from Assemblymember Mia Bonta and sponsored by the California Nurses Association, went the other way. Healthcare IT News reports it restricts AI to advisory roles in clinical care and prevents AI from performing licensed clinical functions, while exempting AI used for clinical documentation and for communications that don’t require professional judgment. The Transparency Coalition adds that it treats businesses offering health care chatbots as providers under California’s medical confidentiality law.

Newsom also signed SB 503 from Sen. Akilah Weber Pierson, which requires developers and deployers to identify clinical decision support systems with known risks of biased impact and make reasonable efforts to mitigate them, per the same report.

Our read: the line Newsom drew is about setting, not principle. AI may not act as the clinician where licensed clinicians already work. SB 903 went after the consumer side, where a companion chatbot can be marketed as something like therapy without a clinician anywhere in the loop. That was the part he would not sign.

Workers got a 90-day clock

The labour bills fared better. SB 951, from Sen. Eloise Gómez Reyes, requires 90 days’ advance notice before technological changes that affect 25% or more of a workforce, according to the Transparency Coalition. As introduced in February, Reyes’ office said notice would go to affected workers, local governments and the state Employment Development Department, triggered at 25 employees or 25% of the workforce, whichever is less. We could not confirm the final thresholds against the chaptered text, and they may have changed in amendments.

SB 947 from Sen. Jerry McNerney bars employers from relying solely on automated decision systems for discipline or termination, and AB 1883 from Assemblymember Isaac Bryan restricts workplace AI surveillance tools from collecting neural data or recognizing emotions, per Stauss. Newsom’s framing, as quoted by the Transparency Coalition: “AI should expand opportunity, not come at the expense of workers.”

What’s happened since: dates, the veto reasoning, and a gap

The calendar is now clearer. AB 1979 and SB 503 take effect Jan. 1, 2027, and SB 947, the automated-discipline bill, becomes operative July 1, 2027, per the California Health Care Foundation, which counts all three as health care AI laws. Professional licensing boards, not a new AI regulator, will enforce AB 1979.

The foundation also summarizes the two vetoes. On SB 903, Newsom argued its definitions and scope were broad enough to catch legitimate clinical tools never meant to replace psychotherapy. On AB 2575, he objected that its “solely based on” evidentiary standard was impractical and that the labor commissioner lacks the medical expertise to judge clinical standards of care. Both are the foundation’s paraphrases; we still have not read the veto PDFs ourselves.

The only AI policy machinery still running in Sacramento points elsewhere. Newsom’s Sept. 18 executive order speeds up the SB 813 and AB 1405 auditor framework and asks experts for recommendations within two months, and his office named those experts on Sept. 23. That puts a report due around mid-November, by our count. The foundation notes the order targets frontier general-purpose models, not clinical tools, so it doesn’t reach the therapy-app question either.

For the wider set of bills Newsom signed and vetoed, see our rundown of California’s 2026 AI laws, and our state AI law map for effective dates beyond California.

The kids’ bills and the auditors came earlier

Several bills that some year-end roundups bundle into the Sept. 30 haul were signed weeks before. SB 1119, from Padilla with Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, requires independent child safety audits and annual risk assessments for companion chatbots, plus crisis protocols, parental controls and safety notifications. Padilla’s SB 867 regulates toys that are companion chatbots. Both were signed Sept. 10.

SB 813 and AB 1405, which set up oversight of independent verification organizations and a state registry for AI auditors with standards for independence, transparency and integrity, were signed Sept. 9. Stauss lists all four with a Sept. 30 date; the governor’s own releases say otherwise, and we go with the governor.

Counts vary with the definition of an AI bill. Stauss tallies 15 AI bills signed and two vetoed in its year-end review. The Transparency Coalition counted 11 signed on the final day alone and nearly 30 AI-related measures enacted across the session. AB 1979 and SB 503 take effect Jan. 1, 2027, and SB 947 becomes operative July 1, 2027, per the California Health Care Foundation; compliance teams should still check each chaptered bill.

Padilla’s sentence, the one about AI not being “the professional itself,” is now roughly the law inside a California hospital. On a phone, in an app that calls itself a listener, it is a question for the next governor.

// How we reported this

Researched and drafted with AI tools, fact-checked against primary sources and approved by a human editor before publishing. Prices and dates are as of publication. Our standards →

Sources
  1. sd18.senate.ca.gov — put it in an Aug. 31 release
  2. gov.ca.gov — vetoed it
  3. chcf.org — the California Health Care Foundation's Oct. 8 review
  4. calmatters.digitaldemocracy.org — CalMatters' Digital Democracy summary
  5. sd18.senate.ca.gov — an Oct. 1 statement
  6. staussfirm.com — a tally by the privacy law firm Stauss
Policy Editor Felix Strauss

Felix Strauss covers tech policy and regulation for prompt/power, from Brussels and Ottawa to Washington and Sacramento. He reads the 400-page regulation so you don't have to, and highlights the one sentence that actually matters.

All stories by Felix →

Latest from prompt/power

  1. Pay yearly or monthly? What you save on AI subscriptions in CanadaNov 19
  2. IBM, NVIDIA and Fermilab plan a quantum test bed. Here’s what each brings, and what’s still missingNov 19
  3. GPT-6.2 Astra is OpenAI’s first released model rated ‘High’ for AI self-improvement. What that means, and what it costsNov 18
  4. OpenAI’s free CSAM classifier misses half the infant images in its own test set. What the model card showsNov 18
  5. Anthropic’s last filing before pricing: Nasdaq approval, a US$6.5-billion stock-pay bill and a redrawn lock-upNov 18

One response to “Newsom Wouldn’t Ban AI Therapists. He Signed the Clinic Version.”

Leave a Reply

Your email address will not be published. Required fields are marked *