California’s legislature has approved Senate Bill 903, which stops AI companion chatbots from presenting themselves as therapists. The measure passed the Assembly 71-4 after clearing the Senate unanimously. It now sits with Governor Gavin Newsom, who has not said whether he will sign it.
Senator Steve Padilla wrote the bill, which targets consumer apps such as Therabot and Wysa. Under its terms, digital tools of that kind cannot offer, advertise or describe their services as therapy or psychotherapy inside the state.
For licensed clinicians, the rules still allow automation but with hard limits. AI may draft notes or help with scheduling, and any automated transcription of a session requires explicit patient consent. Algorithms cannot independently interact with patients, diagnose conditions, judge emotions or build treatment plans without direct oversight from a licensed practitioner.
The push reflects a sharp shortage of mental health workers, with nearly a third of Californians living in areas short of providers, and growing unease about chatbots that miss nonverbal cues or hallucinate harmful advice. Padilla framed the bill as keeping AI a tool in professional hands rather than letting it become the professional itself.
