Colorado, Maine, Rhode Island, Tennessee and Vermont enacted laws in 2026 restricting artificial intelligence chatbots from providing therapy, joining Illinois and Nevada, which passed similar bans in 2025. The wave of state legislation reflects growing alarm among lawmakers and clinicians that millions of people are turning to general-purpose AI chatbots for mental health support without any licensing, oversight or safety guarantees.
What the New Laws Require
Colorado’s HB 1195, taking effect August 12, 2026, prohibits the use of AI therapy chatbots and places new restrictions on how licensed mental health professionals themselves can use AI tools in treatment. Maine’s LD 2082, effective July 29, 2026, sponsored by lawmakers Kuhn and Pierce, treats the offering of therapy services via AI without a licensed human professional as an unfair trade practice, opening the door to consumer-protection enforcement rather than just medical licensing boards. Tennessee’s SB 1580, effective July 1, 2026, bars developers from building or marketing AI systems that represent themselves as qualified mental health professionals. Vermont’s H 816 took effect June 17, 2026, requiring that mental health services be delivered by licensed professionals rather than AI systems. Rhode Island’s companion bills, H 7349 and S 2197, will not take effect until January 1, 2027, but go further by barring even licensed providers from letting AI make independent therapeutic decisions on their behalf.
How This Started
Illinois and Nevada set the template in 2025, becoming the first states to bar AI from delivering therapy to the public, largely by invoking their existing authority to license psychologists, psychiatrists and counselors. That legal approach, treating unlicensed AI therapy as the unauthorized practice of a licensed profession, has since been copied and adapted across the five new states, according to tracking by the Transparency Coalition, an AI policy advocacy group. The push accelerated after a string of publicized cases in which people in mental health crises confided in general-purpose chatbots such as Character.AI and ChatGPT, sometimes with tragic outcomes that drew lawsuits and national attention.
A Patchwork Rather Than a Ban
Not every state is moving toward prohibition. Utah, New York, California and Nebraska have instead opted to regulate rather than ban companion and mental-health-adjacent chatbots. Utah requires mental health chatbots to disclose that they are AI, both up front and whenever a user asks. New York’s law, aimed at companion apps designed to act like a friend or romantic partner, requires that disclosure at the start of a conversation and again every three hours, along with a requirement to route users expressing self-harm ideation to a crisis hotline. The result is an increasingly fragmented legal landscape where the rules a chatbot must follow depend entirely on the user’s state. Compliance teams at companies operating nationwide products now have to track a shifting map of outright bans, disclosure mandates and crisis-referral requirements, a burden that smaller mental health app developers say is harder to absorb than it is for larger, better-resourced competitors.
Industry and Clinical Reaction
Mental health professional associations have broadly supported the restrictions, arguing that AI systems built on general-purpose language models are not clinically validated and can miss or mishandle warning signs of suicidal ideation, an argument echoed in past American Psychological Association advisories on chatbot use for mental health. Digital health companies, meanwhile, have warned that blanket restrictions could sweep up lower-risk tools, such as AI-assisted intake screening or CBT-based self-help apps that operate as adjuncts to licensed care rather than replacements for it. Some venture investors covering the sector have noted that funding is shifting away from AI-as-therapist products and toward tools that support human clinicians instead, a trend documented in 2026 digital health funding data. That shift has already begun reshaping product roadmaps industry-wide, with several digital mental health platforms publicly repositioning their AI features as intake triage or clinician-support tools rather than anything resembling therapy itself, in an apparent effort to stay clear of the growing list of state restrictions.
What It Means Going Forward
With seven states now restricting AI therapy in some form and more legislation pending, companies operating AI mental health products face a state-by-state compliance puzzle that is likely to intensify as 2026 continues. Legal experts expect additional states to introduce similar bills in their next sessions, and federal proposals such as the Senate’s GUARD Act could eventually impose a national baseline. For now, the practical effect is that any AI product marketed as therapeutic must increasingly build in explicit disclosures, crisis-referral logic and, in an expanding list of states, a bright line against claiming to replace a licensed professional altogether.
