Vermont Gov. Phil Scott signed H.816 on June 17, banning AI systems from independently delivering mental health diagnoses, treatment plans, or therapeutic guidance. The law took effect immediately. It’s the clearest signal yet that, with Congress producing nothing comparable, statehouses are writing the operative U.S. rulebook on therapy chatbots one definition at a time.

The Vermont bill, introduced by Rep. Daisy Berbeco (D-Winooski), passed both chambers on a voice vote without opposition, per VTDigger and Seven Days. It carves out clinician use for notes, scheduling, and transcription, and directs the Office of Professional Regulation to deliver recommendations on clinician AI use by Jan. 15, 2027. Scott vetoed the companion bill, H.817, citing a separation-of-powers objection to a budget mandate.

Industry pushed back. The Computer & Communications Industry Association had urged a veto, arguing that H.816’s definitions of “support” and “therapeutic communication” sweep in wellness and productivity apps that no one would call therapy.

That definitional fight is the whole game now, and it’s playing out in parallel across the country.

Per the Transparency Coalition, 78 chatbot bills are alive in 27 states six weeks into the 2026 session. Arizona adjourned sine die on June 13 with three AI bills on Gov. Katie Hobbs’ desk: HB 2311 on chatbot safety and minor protections, HB 2133 on synthetic intimate-image deepfakes, and HB 2592 on state agency AI adoption review. Rhode Island’s House and Senate both approved H 7349, its own therapy chatbot ban, with S 2010 and S 2195 also moving. Pennsylvania’s SB 1090 disclosure bill cleared the Senate 49-1 and sits in House Communications & Technology. California has 30 AI bills in second-chamber committees ahead of the July 2 recess.

The shape is familiar from the post-Dobbs privacy patchwork and the pre-2018 data-breach notification era: when Washington stalls, the fifty-state compliance map becomes the regulation.

Sources

Sources