On September 30, 2026, Governor Newsom vetoed SB 903, a California bill that would have prohibited companies from offering psychotherapy through artificial intelligence (AI) chatbots and limited the extent to which clinicians could rely on AI tools.[1] Proponents argued that the bill kept clinicians involved in patient care, whereas opponents argued that it burdened clinicians by making them consistently review and sign off on AI activity, even if routine.[2] Newsom concluded the bill was overly broad but invited a narrower version and noted the importance of establishing guardrails for AI in healthcare.
SB 903 highlights a national issue: what role should AI play in psychotherapy? Supporters of the AI tools contend AI can expand capacity, enabling clinicians to serve more patients and thereby improving access to healthcare. At the same time, lawmakers have cautioned against over-reliance on AI, and several states have regulated its use. The differing state approaches reflect a lack of consensus on how AI use in psychotherapy should be regulated. The result is an evolving legal landscape, with new legislation expected in 2027.
Lawmakers Encourage Use of AI, But Recognize that Some Guardrails Are Appropriate
Artificial intelligence has swept the nation, transforming how businesses, including healthcare providers, operate day to day. As of February 2026, around half of U.S. adults report using AI chatbots, with a quarter of adults using chatbots daily.[3] The Trump administration has encouraged AI development, with President Trump issuing Executive Order 14365 on December 11, 2025, which stated that the United States would promote “a minimally burdensome national policy framework for AI” and oppose “cumbersome regulation.”[4]
In the healthcare context, states generally permit AI use, but several lawmakers have called for some degree of AI regulation. Several states, for example, have enacted laws that prohibit health plans from using AI to make clinical decisions without clinician involvement,[5] with similar bills pending at the federal level.[6] Other states make clear that, when AI is used, healthcare providers retain independent professional judgment over their patients’ care.[7] Some states have also passed AI disclosure or consent laws, requiring healthcare providers or health plans to make disclosures to or inform patients when AI is used in connection with their care or medical claims.[8]
States Differ in How They Regulate AI in the Context of Psychotherapy
Among states that have enacted or considered specific AI laws concerning psychotherapy, the focus has been on (i) regulation of AI mental health chatbots and (ii) the extent to which a psychotherapy provider may use AI for clinical purposes. Among states that have enacted legislation, the terms and scope of those laws differ.
AI Mental Health Chatbots: Nevada prohibits AI providers from representing that AI can provide psychotherapy and from making AI mental health chatbots available.[9] Colorado similarly prohibits AI providers from representing AI as a therapist, but does not ban making AI mental health chatbots available.[10] Instead, Colorado will prohibit health plans from paying for AI psychotherapy services.[11] Tennessee also prohibits representing AI as a therapist, but without the further prohibitions that Nevada and Colorado have.[12] Illinois, Maine, Rhode Island, and Vermont approach the issue by requiring psychotherapy to come from a qualified professional.[13] Although California’s SB 903, which would have prohibited offering psychotherapy through AI mental health chatbots, was vetoed,[14] AB 1979 will prohibit health facilities with limited exceptions from using AI to independently perform the functions of a licensed professional.[15] Utah takes a different approach, expressly permitting AI mental health chatbots, provided the chatbot discloses it is AI.[16] The federal government has similarly announced a pilot program with an AI device to be used for certain Medicare treatments.[17]
Extent of Permitted Clinical AI Use: State statutes are not consistent on how much a mental health therapist may rely on AI for actual clinical work. Nevada, for example, generally bars a psychotherapist from using AI beyond administrative tasks.[18] Rhode Island’s statute provides that therapists with limited exceptions may not use AI to “[d]etermine therapeutic recommendations or treatment plans.”[19]
On the other hand, Colorado and Maine prohibit clinicians from letting AI make independent clinical decisions, but allow AI to engage in certain functions, including therapeutic recommendations and treatment plans, so long as the clinician reviews and approves the output.[20] The Illinois law is similar, except it prohibits a psychotherapist from using AI to “detect emotions or mental states,” with no review-and-approval exception.[21]
California’s SB 903 would have aligned most closely with Colorado and Maine, since it would have permitted AI to make therapeutic decisions with a licensed professional’s “review and approval,” provided the clinician “uses the [AI] in a manner that is clinically appropriate and compliant with this chapter.”[22] Governor Newsom, however, stated that even with the review-and-approval exception, SB 903 went too far because it would “limit a clinician’s use of tools” including by “requiring routine screening determinations to receive direct approval.”[23]
Takeaways and Next Steps
AI is being used in healthcare more and more, but lawmakers have signaled that some regulation is necessary to ensure proper care. A number of state laws address AI in psychotherapy, but those laws are not consistent. Governor Newsom’s veto of SB 903 underscores the lack of consensus and signals further changes to the legal landscape.
Providers of psychotherapy services should consider monitoring legislative developments in their jurisdictions, including tracking proposed bills and newly enacted laws. The legal landscape is evolving rapidly, and the requirements governing AI-assisted psychotherapy could change substantially over the next year.
In California, the fight over SB 903 is not over. After Governor Newsom vetoed the bill, Senator Padilla, the bill’s author, stated that he would reintroduce the legislation in the coming legislative session to prevent AI from providing psychotherapy.[24] The 2027-28 legislative session will convene on December 7, 2026, so expect to see legislation introduced again in the near future.
FOOTNOTES
[1] Governor's Veto Message to Cal. S.B. 903 (Sept. 30, 2026), https://www.gov.ca.gov/wp-content/uploads/2026/09/VETO-msg-SB-903.pdf; S.B. 903, 2025–2026 Leg., Reg. Sess. (Cal. 2026) (vetoed), https://leginfo.legislature.ca.gov/faces/billPdf.xhtml?bill_id=202520260SB903&version=20250SB90394ENR.
[2] Assemb. Comm. on Privacy & Consumer Prot., Analysis of S.B. 903, 2025–2026 Leg., Reg. Sess., at 11-12 (Cal. July 1, 2026), https://apcp.assembly.ca.gov/system/files/2026-07/sb-903-padilla-apcp-analysis.pdf.
[3] Jeffrey Gottfried et al., Americans and AI 2026: Chatbots, Smart Devices and Views on Impact, Pew Rsch. Ctr. (June 17, 2026), https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/.
[4] Exec. Order No. 14365, 90 Fed. Reg. 58,499 (Dec. 16, 2025), https://www.govinfo.gov/content/pkg/FR-2025-12-16/pdf/2025-23092.pdf.
[5] Alabama, Act No. 2026-589, § 1(b)(3), https://arc-sos.state.al.us/ucp/L2123253.AI1.pdf; California, S.B. 1120, §§ 1(k), 2(j), https://leginfo.legislature.ca.gov/faces/billPdf.xhtml?bill_id=202320240SB1120&version=20230SB112090CHP; Colorado, H.B. 26-1139, § 2 (eff. Jan. 1, 2027), https://leg.colorado.gov/laws/session-laws/HB26-1139/325/download; Georgia, Act No. 411, § 1(c) (eff. Jan. 1, 2027), https://gov.georgia.gov/document/2026-signed-legislation/sb-444/download; Indiana, H.B. 1271, § 6, https://iga.in.gov/pdf-documents/124/2026/house/bills/HB1271/HB1271.07.ENRS.pdf; Iowa, H.F. 2635, § 2, https://www.legis.iowa.gov/docs/publications/LGE/91/HF2635.pdf; Nebraska, L.B. 77, § 12, https://nebraskalegislature.gov/FloorDocs/109/PDF/Slip/LB77.pdf.
[6] H.R. 10210, §§ 2(b), 3(b) (Doctors Not AI Act of 2026) (introduced Sept. 1, 2026) (group health plans), https://www.congress.gov/119/bills/hr10210/BILLS-119hr10210ih.pdf; H.R. 9734 (Protecting Patients from Automated Denials Act) (introduced July 16, 2026) (Medicare Advantage plans), https://www.congress.gov/119/bills/hr9734/BILLS-119hr9734ih.pdf.
[7] E.g., California, A.B. 1979, § 1 (eff. Jan. 1, 2027), https://leginfo.legislature.ca.gov/faces/billPdf.xhtml?bill_id=202520260AB1979&version=20250AB197989CHP.
[8] E.g., Alabama, Act No. 2026-589, § 1(c); California, S.B. 1120, §§ 1(k), 2(j); Indiana, H.B. 1271, § 6; Nebraska, L.B. 77, § 12.
[9] Nevada, A.B. 406, § 7, https://archive.leg.state.nv.us/Session/83rd2025/Bills/AB/AB406_EN.pdf.
[10] Colorado, H.B. 26-1195, § 3, https://leg.colorado.gov/laws/session-laws/HB26-1195/358/download.
[11] Colorado, H.B. 26-1139, § 2(6) (eff. Jan. 1, 2027).
[12] Tennessee, S.B. 1580, § 1, https://publications.tnsosfiles.com/acts/114/pub/pc0647.pdf.
[13] Illinois, Pub. Act 104-0054, § 20(a), https://www.ilga.gov/Documents/Legislation/PublicActs/104/PDF/104-0054.pdf; Maine, L.D. 2082, https://legislature.maine.gov/legis/bills/getPDF.asp?paper=HP1397&item=3&snum=132; Rhode Island, S. 2197, § 1, https://webserver.rilegislature.gov/PublicLaws/law26/law26374.htm; Vermont, H. 816, https://legislature.vermont.gov/Documents/2026/Docs/ACTS/ACT156/ACT156%20As%20Enacted.pdf.
[14] California, S.B. 903, § 1 (vetoed Sept. 30, 2026).
[15] California, A.B. 1979, § 1 (eff. Jan. 1, 2027).
[16] Utah, H.B. 452, § 5, https://le.utah.gov/Session/2025/bills/enrolled/HB0452.pdf.
[17] U.S. Food & Drug Administration, Participants Selected for TEMPO for Digital Health Devices Pilot, FDA.gov (July 22, 2026), https://www.fda.gov/medical-devices/digital-health-center-excellence/participants-selected-tempo-digital-health-devices-pilot.
[18] Nevada, A.B. 406, § 8.
[19] Rhode Island, S. 2197, § 1.
[20] Colorado, H.B. 26-1195, § 2(5)(b); Maine, L.D. 2082.
[21] Illinois, Pub. Act 104-0054, § 20(b).
[22] California, S.B. 903, § 1 (vetoed Sept. 30, 2026).
[23] Governor's Veto Message to Cal. S.B. 903 (Sept. 30, 2026).
[24] Press Release, Office of Cal. State Sen. Steve Padilla, Governor Newsom Allows Dangerous Chatbots to Provide Therapy (Oct. 1, 2026), https://sd18.senate.ca.gov/news/governor-newsom-allows-dangerous-chatbots-provide-therapy.