AI Regulation Tracker / Regulation enacted
Rhode Island Just Drew a Hard Line Around AI in Mental Health Care. Here Is What Clinicians Owe
Two new Rhode Island laws require AI chatbots to flag suicide risk and refer users to crisis help, and bar AI from making therapeutic decisions without licensed human oversight and patient consent. Here is what is binding and the first move for any practice using AI.
Two laws, one boundary
Rhode Island just told AI where it is allowed to stand in a therapy room. On June 23, 2026, Governor Dan McKee signed two bills, sponsored by Senator Lori Urso and Representative Tina Spears, that together draw a hard line between AI as a support tool and AI as a substitute for a licensed clinician. The bills were part of a broader Senate health care package, and the sponsors were blunt about why they wrote them. "Chatbots aren't human and they do not live up to the professional standards of a trained therapist," Representative Spears said in the General Assembly's June 23 announcement.
The laws split the work in two. One governs any AI chatbot a distressed person might reach. The other governs AI used inside licensed mental health care.
The chatbot safety law
The first bill, 2026-S 2195Aaa and its House companion 2026-H 7350Aaa, applies to AI chatbots broadly. It requires developers to build protocols that protect vulnerable users who express suicidal ideation, a desire to self-harm, or a desire to harm others while interacting with the system. When the chatbot detects any of those expressions, it must refer the user to a crisis service provider. Developers also have to file annual reports with the Attorney General's office showing how many times their safety protocols activated and related metrics.
The law adds a disclosure most users would assume already existed. A chatbot must clearly and conspicuously tell users that they are not communicating with a human, both at the start of a session and at least once every three hours during an ongoing one. The bill's sponsors tied this directly to documented tragedies, citing testimony from parents whose children died by suicide after extended conversations with AI companions that failed to point them toward help.
The mental-health-care law
The second bill, 2026-S 2197Aaa and its House companion 2026-H 7349A, is the one every clinician should read closely. It sets guidelines for AI used in mental health care with three load-bearing rules.
First, consent. A patient, or their parent or guardian, must give fully informed consent before an AI companion model can be used to assist in a therapy or psychotherapy session. Second, human authority. Licensed mental health professionals, not AI, must be the ones making therapeutic decisions, determining therapeutic recommendations, and creating treatment plans. Third, confidentiality. AI tools used in mental health care, including those that handle administrative tasks, note-taking, supplementary support, or therapeutic communication, must meet the same confidentiality standards that apply to every other record a licensed professional keeps.
The law also closes a marketing gap. It prohibits companies and individuals from offering or advertising therapy or psychotherapy services unless a licensed professional conducts them. That is the provision that functions as a ban on AI marketed as a stand-in therapist.
What this means for a working clinician
None of this bars a practice from using AI. It tells the practice where the human has to stay in the chair. A scribe that drafts a progress note is support. A model that quietly shapes a treatment plan is the thing the law now forbids without a licensed professional owning the decision. The distinction the statute draws, between AI that assists and AI that decides, is exactly the line a careful clinician would draw anyway. Rhode Island just made it enforceable.
In written testimony supporting the care bill, Family Service of Rhode Island warned that "an over-dependence on AI to guide clinical treatment can result in errant decision-making and harmful outcomes in client care." The law turns that warning into a rule.
The operating picture
For a Rhode Island practice the work is concrete. Build an inventory of every AI tool that touches a patient, from intake to notes to any companion model. Confirm a licensed clinician signs every therapeutic recommendation and treatment plan. Add an AI-use consent step to intake so the informed-consent requirement is documented, not assumed. And confirm any vendor handling clinical content meets the confidentiality standard the law now imports. The practices that already treat AI as an assistant rather than a decision-maker will find they are most of the way there.
Frequently Asked Questions
When did Rhode Island's AI mental health laws take effect?
Both bills were signed into law on June 23, 2026. One sets safety requirements for AI chatbots generally, and the other sets rules for AI used in mental health care.
Does the law ban AI in therapy?
No. It bans AI from making therapeutic decisions and from being marketed as a stand-in therapist. AI may still assist with tasks like note-taking and administrative support, though a licensed professional must make therapeutic decisions and treatment plans, and the patient or guardian must give informed consent before an AI companion model assists in a session.
What does the chatbot safety law require?
Under 2026-S 2195Aaa, AI chatbot developers must build protocols to detect expressions of suicidal ideation, self-harm, or intent to harm others, refer affected users to a crisis service provider, and file annual reports with the Attorney General on how often those protocols activate. Chatbots must also tell users they are not human at the start of a session and at least once every three hours during a continuing one.
Do AI note-takers fall under the confidentiality rule?
Yes. The mental-health-care law extends the confidentiality standards that apply to a licensed professional's records to AI tools used in mental health care, including those used for administrative tasks, note-taking, supplementary support, and therapeutic communication.
Sponsored Training
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.