Vermont mandates therapist oversight for AI mental health tools
Vermont has enacted legislation requiring human therapists to review and approve any AI-generated mental health advice, adding compliance burdens and litigation risks for AI developers in a fragmented US regulatory landscape.
Vermont enacted H.816 (Act 156) into law on June 17, 2026, barring companies from providing AI-driven mental health services unless a licensed therapist reviews and approves the output. The legislation is part of a growing wave of state-level tech regulation targeting the healthcare sector.
The law explicitly aims to shield users from psychological harm, including suicide, by ensuring AI systems do not operate independently in a clinical capacity. Under the new rules, corporations can only offer these tools if a mental health professional operating within their scope of practice signs off on the AI's recommendations.
Compliance and operational risks
For AI developers and digital health companies, the Vermont mandate introduces a direct threat to the scalability of their products. The core commercial appeal of generative AI is its ability to provide cheap, 24/7 access, which human oversight inherently limits.
There is also a pronounced risk that this oversight mechanism fails in practice. Therapists tasked with reviewing AI outputs may simply rubber-stamp the algorithm's advice due to time constraints or an overreliance on machine-generated text, leaving companies legally exposed if a user suffers harm.
A fragmented regulatory landscape
Vermont's law is not an outlier but rather the latest in a series of state-level regulations, following similar frameworks in Illinois, Nevada, and Utah. This geographic patchwork complicates market entry for AI providers, forcing them to navigate varying state standards rather than a unified federal rule.
Congress has so far failed to pass an overarching federal AI law. If federal legislation eventually emerges, it could preempt these state laws, triggering a wave of legal battles over jurisdiction and compliance.
Until then, AI makers face significant legal uncertainty. These state statutes have not yet been tested in court, but the litigation threat is real. OpenAI faced a high-profile lawsuit last year alleging inadequate safeguards in its cognitive advisement features.
The market size for these tools is substantial. ChatGPT alone boasts 900 million weekly active users, with mental health queries ranking among the top uses for general-purpose AI models like GPT-5, Claude, and Gemini. However, developers remain exposed because these general-purpose systems lack the robust clinical capabilities of human therapists, while specialized, purpose-built mental health AI remains largely in the testing phase.
