Newsom Vetoes California Bill Barring AI Chatbots From Therapy

Gov. Gavin Newsom vetoed Senate Bill 903 on Wednesday, rejecting legislation that would have barred companies from offering or advertising chatbot-delivered therapy in California and drawing a sharp rebuke from the senator who wrote it. The bill had cleared the Senate unanimously and the Assembly by a vote of 71 to 4.

The measure, by Sen. Steve Padilla (D-San Diego), would have required that therapy and psychotherapy be delivered by licensed professionals. It would have prohibited individuals and companies — including those using artificial intelligence — from offering or advertising "therapy" when the service was provided by a companion chatbot, confined AI to administrative and supplementary roles under a licensed clinician's responsibility, and required notice before AI was used to record, transcribe, triage, or screen clients.

In his veto message, Newsom called the bill "overly broad" and said it "would drastically limit a clinician's use of tools that benefit the delivery of care today, including by requiring routine screening determinations to receive direct approval."

What the bill would have added to state law

SB 903 would have drawn a line that California's existing licensing rules leave blurry: the difference between a licensed professional using software and software acting as the professional. Under the bill, AI could not independently interact with clients in psychotherapy, make therapeutic decisions, detect a client's emotional state, or generate treatment plans without review by a licensed professional. Enforcement would have gone to the state's healing arts boards.

A policy analysis prepared for the Assembly Committee on Privacy and Consumer Protection noted that a subset of consumer products — it named Woebot and Wysa — are purpose-built to deliver cognitive behavioral therapy, an established clinical treatment, rather than the open-ended conversation of entertainment chatbots. The analysis put the question squarely before lawmakers: whether marketing such a product as therapy should require the involvement of someone licensed to provide it.

Padilla's office said thousands of licensed mental health professionals signed a letter urging the governor to sign the measure.

Padilla: the veto leaves "unlicensed algorithms"

Padilla said the decision contradicts the governor's own recent record. "Earlier this month, I stood alongside Governor Newsom and other leaders as he signed the nation's strongest protections for children interacting with chatbots into law, which is what makes this so much more disappointing," he said in a 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."

Padilla has authored several of the state's AI statutes, including first-in-the-nation guardrails for companion chatbots and a moratorium on AI-enabled toys. He said he plans to reintroduce the therapy measure in the coming legislative session.

His office framed the issue as urgent, pointing to federal efforts to bring AI agents that offer therapy and prescribe medication into Medicare, with technology companies positioned to profit from the rollout.

A batch of health-AI actions, signed and refused

The veto arrived in the same round of bill actions in which Newsom signed 22 of 25 privacy and artificial-intelligence measures sent to him, including new obligations for companies whose products children use. Two of the signed bills touch clinical care directly. AB 1979 bars health facilities, clinics, and physician offices from deploying AI to independently perform clinical functions that state law reserves for licensed people, and extends confidentiality protections to records accessed through direct-to-consumer health chatbots. SB 503 requires developers and users of clinical decision support systems to identify tools that carry a foreseeable risk of biased outcomes and to monitor them after deployment.

Newsom also vetoed AB 2575, which would have shielded health care workers from retaliation when they override an AI recommendation. The California Nurses Association criticized that decision, with president Sandy Reding saying a nurse who spots a faulty algorithm and acts on it "can still be disciplined or fired by her employer."

Where the rules go from here

For Californians seeking help for depression, anxiety disorders, or a substance use disorder, the practical effect of the veto is narrow. A vetoed bill is not law, so no new AI-specific sign-off requirement now applies to clinicians who use screening, intake, or note-writing tools. Existing duties around supervision, informed consent, confidentiality, and truthful advertising are unchanged, and California law already limits the practice of psychotherapy to licensed providers — just without SB 903's AI-specific wording.

The veto does not settle the underlying question. It moves the argument to how narrowly any future rules should be written: whether a ban targets chatbot-delivered therapy specifically, or sweeps in the routine tools that clinics have adopted, including the telehealth services and app-based products that expanded during the pandemic. Newsom left the door open to revisiting the issue, saying in prior remarks that the state has to act in the absence of federal leadership. Padilla has said he will try again.

Whether other states follow is an open question. Several have already acted on AI therapy through their own licensing boards, and the absence of a federal standard means the map is being drawn state by state — with a veto in the country's most populous state now part of it.