A California bill that would bar AI chatbots from acting as unsupervised therapists cleared the Assembly Appropriations Committee 13-0 on August 13, moving it one step closer to a full Assembly floor vote, according to Decrypt. Senate Bill 903, dubbed the Wellness and Oversight for Psychological Resources Act, was introduced by state Senator Steve Padilla on January 21 and has advanced through the legislature with near-unanimous support, passing the full Senate 39-0 in May and clearing two Assembly policy committees by margins of 17-0 and 14-1.
The bill would prohibit companion chatbots from independently offering psychotherapy, require clinician oversight and patient consent for AI-driven mental health tools, and limit AI systems to a support-only role in therapeutic settings. It carves out exemptions for peer support programs, religious counseling, academic research, and tools that already have federal medical approval.
"Not fit to take over the job"
Padilla has framed the bill as a response to a technology moving faster than the safeguards around it. "AI algorithms are not fit to take over the job of human therapists," he said, according to Decrypt's reporting. "We must act and place guardrails." The official bill text on the California Legislature's website lays out the licensing and oversight requirements in full.
The scale of the underlying trend gives the bill's backers their strongest argument: a June survey by the American Psychological Association found that 77% of psychologists reported patients using AI tools for mental health support, often outside any clinical relationship. Researchers at King's College London have separately documented cases where chatbots reinforced users' delusions through excessive agreement rather than pushing back, a failure mode that has no equivalent safeguard in a licensed human therapist.
California would join a small but growing list of states
Related: Google's Cheap New Gemini Model Targets Crypto's Trading Bots
California isn't moving first. Utah's HB 452 took effect May 7, 2025, requiring AI mental-health tools to clearly disclose they aren't human before first use, after seven days of inactivity, and whenever a user asks, while restricting how user data can be sold or shared. Nevada's AB 406 went further, banning AI systems from providing professional mental or behavioral health services outright as of July 1, 2025, and outlawing marketing terms like "AI therapist" or "virtual psychotherapist" without clinician supervision. Illinois followed on August 4, 2025, empowering its Department of Financial and Professional Regulation to fine violators up to $10,000 per incident.
Not everyone backing faster AI adoption in healthcare is on board. Robert Boykin of the industry group TechNet warned that SB 903 "still puts a clinician bottleneck in front of intake and screening tools that help patients reach care faster," pointing to California's existing shortage of behavioral healthcare workers as a reason to keep AI-assisted triage tools unencumbered. With bills already introduced in Florida, Massachusetts, New Hampshire, New York, Ohio and Pennsylvania, California's decision on SB 903 is likely to become a reference point for a wave of similar legislation moving through statehouses over the next year.
