With no federal framework yet governing AI chatbots that people turn to for mental health support, a growing patchwork of state laws is filling the gap in 2026, even as the Food and Drug Administration continues weighing how, or whether, to regulate the technology nationally. Illinois and Nevada have already gone furthest, barring AI from independently delivering therapy, while Utah has taken a lighter-touch approach built around disclosure rather than prohibition, illustrating just how differently states are choosing to handle a technology millions of people now use daily for emotional support.
Illinois and Nevada Draw a Hard Line
Illinois’s Wellness and Oversight for Psychological Resources Act, in force since August 2025, forbids AI systems from delivering therapy on their own or being marketed as a therapist unless a licensed professional remains actively in charge of care, with fines of up to $10,000 per violation for companies that cross the line. Nevada’s Assembly Bill 406, live since July 2025, similarly bars AI from standing in for a counselor or psychologist, and its reach extends into schools, where AI-based counseling tools have proliferated amid shortages of school psychologists. Both states leave room for AI to handle administrative tasks like scheduling and paperwork, but draw a firm boundary around anything resembling clinical care delivered directly to a patient.
Utah Chooses Disclosure Over Prohibition
Utah took a different path with House Bill 452, which does not ban mental health chatbots but instead requires them to plainly state that they are software rather than a human clinician, restricts how they can be advertised, and limits what companies can do with the sensitive personal data users share during emotionally vulnerable conversations. Policy analysts tracking the state-by-state landscape describe Utah’s approach as one of at least three distinct regulatory models now emerging nationally: outright prohibition in Illinois and Nevada, disclosure-with-safe-harbor in Utah, and a third model in states such as California and New York that require crisis-detection safeguards backed by the threat of private lawsuits if a chatbot fails to flag a user in danger.
Why the Federal Government Is Still on the Sidelines
The FDA’s Digital Health Advisory Committee held a public meeting on November 6, 2025, specifically to confront the risks of therapy chatbots built on large language models, with the public comment docket closing on December 8, 2025. FDA staff are expected to synthesize that input into a discussion paper, with industry watchers anticipating voluntary pilot pathways rather than binding rules as an initial step. Notably, while the FDA has cleared more than 1,200 AI-based digital health devices for marketing, none of them has been indicated specifically for treating mental health conditions, leaving the fastest-growing corner of consumer AI mental health use almost entirely outside existing device regulation.
The Scale of the Problem Driving Urgency
Survey research published in JAMA earlier in 2026 found that millions of people are already turning to general-purpose AI chatbots for mental health support, often because therapist appointments are unaffordable, unavailable, or come with long waitlists, rather than because a purpose-built therapy app was their first choice. That reality has put pressure on lawmakers to act quickly even without federal guidance, particularly after several high-profile cases in which vulnerable users, including minors, received harmful or inadequate responses from chatbots during moments of psychiatric crisis, incidents that have driven much of the state-level legislative momentum in 2025 and 2026.
Two Camps, Two Very Different Bets
Advocates for the stricter Illinois and Nevada model argue that AI language models remain fundamentally unable to exercise clinical judgment in high-stakes moments and that allowing them to operate as unsupervised therapists risks real harm at a scale regulators cannot easily monitor after the fact. Proponents of Utah’s lighter model counter that outright bans risk pushing vulnerable users toward unregulated, offshore, or underground AI tools with even fewer safeguards, and argue that transparency and data protections, paired with continued clinician oversight of higher-risk cases, better balance access against safety given how many people already rely on these tools daily.
What to Watch Through the Rest of 2026
With the FDA’s Digital Health Center of Excellence expected to publish updated generative AI guidance later this year, industry groups and state legislators alike are watching closely for signals about whether Washington will eventually preempt the current state patchwork or simply codify a version of it nationally. In the meantime, companies building AI mental health products face a fragmented compliance landscape in which the same chatbot may be lawful in Utah, restricted in California, and effectively prohibited in Illinois and Nevada, a situation legal analysts expect to persist at least through the end of 2026 as more states introduce their own versions of chatbot oversight legislation.
