Vermont's H.816 (Act 156): AI Mental Health Advisement Now Requires Human Therapist Review

Vermont enacted H.816 (Act 156) on June 17, 2026, requiring that any AI-based mental health advisement be reviewed and approved by a professional therapist. The law mirrors a growing wave of state-level AI mental health statutes, but its vague review standards create practical risks for AI builders, clinicians, and users.

What happened

Vermont's H.816 (Act 156), titled "An act relating to regulating the use of artificial intelligence in the provision of mental health services," was signed into law on June 17, 2026. The law stipulates that AI makers cannot provide AI in Vermont that performs mental health advisement unless the AI is overseen by a professional therapist. This is similar to laws in Illinois, Nevada, and Utah. However, unlike Colorado's law which requires real-time synchronous intervention, Vermont's law does not mandate that the therapist be actively monitoring during a chat.

The key concern is that the law does not specify the depth or manner of the review. The act of reviewing and approving is left unspecified, which could lead to therapists rubber-stamping AI outputs due to automation bias or time constraints. The bill's aim, as reported by VTDigger, is to ensure that mental health professionals, not AI chatbots, are the ones making mental health decisions.

Why AI builders should care

For AI builders developing mental health chatbots or advisory tools, this law introduces a human-in-the-loop requirement that directly affects product design. You need to build audit trails that document therapist review and approval. The vague standards also create liability ambiguity: AI makers might argue that harm was the therapist's fault for not catching bad advice, while therapists might blame the AI maker for producing flawed outputs. This finger-pointing could lead to costly litigation.

Additionally, the law applies to general-purpose AI (GPAI) like ChatGPT, Claude, Gemini, and purpose-built AI (PBAI). If your product is used in Vermont for mental health advisement, you must ensure a therapist is in the loop. This may require building features for therapist dashboards, logging, and approval workflows.

Practical implications

Therapists now face new responsibilities and potential malpractice risks. Licensing boards may eventually impose sanctions for inadequate review. The law does not define what constitutes a thorough review, so therapists may need to develop their own standards. Some therapists might even use AI to review AI, which defeats the purpose.

For AI builders, the practical takeaway is to design systems that make human review easy and verifiable. Consider adding features like:

• Clear display of AI recommendations for therapist review
• Time-stamped approval logs
• Alerts for unusual or high-risk advice
• Integration with therapist workflow tools

Caveats

The analysis of Vermont's law is based primarily on Forbes coverage and related articles. The exact statutory language may contain nuances not captured here. The law has not been tested in court, so its interpretation remains uncertain. Federal AI regulation could eventually preempt state laws. Builders should consult legal counsel for compliance.