When a Chatbot Acts as a Therapist: Tennessee’s New Law and AI Mental Health

What happened. In January 2026, Character.AI and Google settled five wrongful-death and injury suits brought by families in Florida, Colorado, New York, and Texas who blamed the companies’ chatbots for their children’s suicides or mental-health crises. The lead case, Garcia v. Character Technologies, was filed in October 2024 by Garcia after the death of her 14-year-old son. In May 2025, a federal judge allowed the product liability and wrongful death claims to proceed, declining at that stage to hold that chatbot output is protected speech. On May 5, 2026, the Pennsylvania Department of State sued Character Technologies in court for the unauthorized practice of medicine, after a bot styled as a “Doctor of psychiatry” claimed to be licensed in Pennsylvania and supplied a fabricated license number. Washington opened a Section 6(b) FTC inquiry into seven chatbot makers in September 2025 and held a Senate hearing on chatbot harms. On July 1, 2026, Tennessee’s SB 1580 took effect, making it illegal to advertise or represent to the public that an AI system is, or can act as, a “qualified mental health professional.”

Why it matters. The striking feature of SB 1580 is how little Tennessee had to build. The bill adds a single new section to the state’s mental health code (Tenn. Code Ann. § 33-1-205) and is enforced through institutions the state already had. It leaves “qualified mental health professional” undefined, the hardest and most litigable term in any such bill, and points instead to the licensure definition already sitting in Title 33, which covers psychiatrists, psychologists, licensed clinical social workers, marital and family therapists, professional counselors, and psychiatric nurses. By declaring a violation an “unfair or deceptive act or practice,” it borrows the remedies of the 1977 Consumer Protection Act: attorney-general enforcement, a private right of action, and a civil penalty set at $5,000 per violation, above the Act’s usual ceiling. It targets the marketing claim rather than the service itself, which sidesteps the two fights that have slowed broader bills: definitional arguments over what counts as therapy, and First Amendment arguments over whether a model’s output is protected speech. A false commercial claim raises neither question.

What happened. Pew’s “Americans and AI 2026,” released in June, found that 49% of US adults now use AI chatbots, up from 33% in 2024 and 23% in 2023, with about a quarter using them daily. One in ten use them for emotional support, and among adults under 30, it is one in five. A Harvard Business Review analysis of online forum posts, published in April 2025, ranked therapy and companionship the single most common use of generative AI, though it draws on self-reported forum discussion rather than a representative sample. The clinical layer confirms the pattern from the other side of the desk: the American Psychological Association’s (APA) 2026 survey of 1,242 licensed psychologists found that 77% have treated patients who used AI for support, and 35% have patients using AI as an additional mental health professional. Those clinicians are near-unanimous on the risk: 97% worry the tools can reinforce dysfunctional beliefs and 89% worry they can inadvertently encourage self-harm, concerns the survey ties to research in Science, where Stanford researchers tested 11 leading models and found they affirmed users 49% more often than human respondents did, and validated the user 47% of the time even when the prompt described deceptive or illegal behavior. A single affirming reply left participants less willing to apologize or repair the conflict, and more confident they had been right.

Why it matters. This is the demand the law is reacting to, and it explains why a single marketing line carries the weight it does. Pew finds that the people most likely to lean on a chatbot for emotional support, under-30s, are also the most skeptical of AI’s trajectory. Adoption is running ahead of both public trust and professional guardrails, and rules are the piece a legislature can move quickly on. Tennessee’s law does not stop the behavior. It polices one specific claim about it, that a chatbot is as good as a professional, and the APA data suggests that is the claim doing measurable harm.

What happened. Tennessee is early, but it is not first. By the time SB 1580 took effect, at least five states had already regulated AI in mental health. Utah HB 452 moved first, effective May 7, 2025, requiring mental-health chatbots to disclose that the user is talking to a machine, barring the sale of user health data, and restricting in-chatbot advertising. Nevada AB 406 took effect July 1, 2025, barring AI outright from providing mental or behavioral healthcare, with penalties up to $15,000. Illinois signed its Wellness and Oversight for Psychological Resources Act in August 2025, banning AI-delivered therapy and treating “AI therapy” or “chatbot counselor” advertising as unlawful. New York’s AI companion law took effect in November 2025, requiring suicide-risk protocols and repeated non-human disclosures. California SB 243 followed in January 2026, adding a private right of action with a $1,000-per-violation minimum. Tennessee, in July 2026, is the sixth. Both Tennessee’s and Illinois’s bills passed unanimously.

Why it matters. Three enforcement models are now in the field: disclosure rules (Utah, New York), practice bans (Nevada, Illinois), and Tennessee’s deceptive-marketing line. The practice bans offer the broadest protection on paper and carry the greatest exposure in court, since they raise both the scope-of-practice question of what counts as therapy and the unsettled First Amendment question about model output. Tennessee’s model raises neither. It is the narrowest of the three and the only one a state can enact without inventing anything, which makes it the easiest template for the next legislature to copy.