A Colorado law restricting how artificial intelligence can be used in psychotherapy took effect on August 12, 2026, making the state one of at least five to pass legislation this year specifically targeting AI’s growing role in mental health care. House Bill 26-1195, the Psychotherapy Artificial Intelligence Restrictions Act, was sponsored by Representatives Gretchen Rydin and Javier Mabrey along with Senators Judy Amabile and Kyle Mullica, and drew 45 cosponsors before Governor Jared Polis signed it into law on June 3, 2026.
What the Law Actually Prohibits
The statute prohibits licensed, certified, or registered mental health professionals from letting an AI system directly engage in therapeutic communication with a client, generate treatment recommendations without a clinician reviewing them first, or attempt to detect a client’s emotions or mental state on its own. It also requires therapists to obtain client consent before using AI to record or transcribe sessions, and it bars anyone who is not a regulated professional from offering psychotherapy services at all, whether human or automated. Violations are classified as unfair trade practices under the Colorado Consumer Protection Act, and the state’s professional licensing boards within the Department of Regulatory Agencies can impose disciplinary action and financial penalties on top of that.
Chatbots Get Their Own Rules
Separately, the law establishes requirements for AI systems that interact with the public around mental health, regardless of whether a licensed professional is involved. Such systems are barred from implying or stating that they are a human provider or are authorized to practice psychotherapy, and any chatbot offered to the public for mental-health-adjacent purposes must implement protocols to respond to a user expressing suicidal ideation or self-harm, including referring the person to a crisis hotline such as the 988 Suicide and Crisis Lifeline. AI use for purely administrative tasks, such as scheduling or basic supplementary support, remains permitted as long as the licensed professional retains full responsibility for all interactions, outputs, and data use.
Part of a Wider Legislative Wave
Colorado’s law joins similar measures passed in Utah, Nevada, Illinois, and Vermont over the past 18 months, reflecting a rapid state-level response to reports of AI chatbots representing themselves as licensed therapists and, in some documented cases, giving advice that clinicians say worsened users’ conditions. According to tracking by the Center for Democracy and Technology and Becker’s Behavioral Health, nearly 126 chatbot-specific bills were introduced across 37 states in 2026 alone, with lawmakers increasingly splitting into two camps: states like Colorado, Illinois, and Vermont that prohibit AI from independently delivering therapy, and states like Utah that instead require disclosure and guardrails while allowing broader use.
Why Lawmakers Moved Now
The legislative push follows a wave of scrutiny of AI companion and chatbot platforms tied to user harm, including congressional attention to products that engaged minors in unsafe conversations. Mental health researchers, including work published by Stanford’s Institute for Human-Centered AI, have documented cases where general-purpose AI chatbots validated users’ delusions or failed to recognize crisis-level distress, while separately, RAND Corporation survey data published in 2026 found that use of AI chatbots for mental health advice among adolescents and young adults had risen sharply year over year. Colorado lawmakers cited concerns that unregulated AI tools were filling gaps left by a shortage of licensed therapists, without the safeguards a human clinician would provide.
Industry and Clinician Reaction Is Split
Digital health companies operating in Colorado have had to audit their products against the new rules, and some have quietly restructured chatbot features to route more decisions through licensed staff to stay compliant. Supporters, including several state psychological associations, argue the law protects vulnerable patients from unproven AI standing in for trained clinicians. Critics, including some AI mental-health startups and free-speech advocates, contend the restrictions could face First Amendment challenges similar to arguments raised against other state laws restricting counselor-client communication, and warn that overly broad rules could choke off AI tools that have shown real, if modest, benefit for mild depression and anxiety symptoms when used as a supplement to traditional care.
What Happens Next
Colorado regulators are still finalizing more detailed implementation rules under a related law, the Colorado AI Act’s Automated Decision-Making Technology provisions and a separate Chatbot Safety Act, both of which are expected to take full effect on January 1, 2027. In the meantime, DORA’s licensing boards have enforcement authority now, and behavioral health providers operating in the state say they are watching closely for the first disciplinary cases to see how aggressively the state intends to apply the new consent and oversight requirements. With similar bills pending in additional states, Colorado’s law is likely to serve as a reference point for legislators elsewhere weighing how far to let AI go inside the therapy room.
Photo: geralt / PIXABAY via Pixabay
