Ban AI “Therapists”: Protect Clients from Serious Harm and Safeguard the Integrity of Mental Health Care

AI tools marketed as ‘AI Therapists’ are not only misleading — they endanger the very individuals they claim to help. These systems often interact with people at moments of profound vulnerability: during episodes of severe depression, heightened anxiety, periods of peak substance use, suicidal ideation, or acute crisis. In such moments, clinical missteps are not benign — they can be life-threatening.

Moreover, these tools risk reinforcing the false belief that professional care is unnecessary — even when symptoms are severe or risk of harm to self or others is imminent. The resulting delay in appropriate treatment is not a matter of convenience — it is a matter of clinical safety.

A growing body of research confirms the seriousness of these concerns:

• A 2025 Stanford University study, led by Jared Moore and Nick Haber from the Institute for Human-Centered Artificial Intelligence (HAI) evaluated the clinical safety of AI therapy chatbots currently marketed to consumers — including Pi (7cups), Noni, and Character.ai’s “Therapist” bot. The researchers found these systems frequently produced stigmatizing, emotionally inappropriate, and clinically unsafe responses, particularly in scenarios involving trauma, suicidality, and delusions. https://stanford.io/3UFCeHn
• A 2024 University College London systematic review published in JMIR Mental Health by Guo, Lai, Thygesen, Farrington, Keen, and Li evaluated the current landscape of large language models (LLMs) applied in mental health care. Reviewing dozens of use cases across AI “therapy” tools, the authors concluded these systems routinely face challenges related to hallucinations, bias, and deeply troubled interpretations of diagnosis. The review found no evidence of clinical effectiveness, and warned that AI therapy applications therefore pose a significant risk to patients/clients when used without licensed clinician oversight. https://jmir.org/2024/mentalhealth/LLM-review
• A 2023 MIT & Yale University study conducted by Zhou, Kang, Klein, and Riedl — explored how users perceived empathy, effectiveness, and trust in interactions with AI-generated therapeutic responses. The study found that participants consistently reported higher levels of emotional safety, perceived empathy, and therapeutic alliance when interacting with responses generated by licensed human therapists. Decades of research validate high therapeutic alliance as one the most important elements necessary to alleviate depression, anxiety, and other acute diagnoses. https://arxiv.org/abs/2306.01872

Together, these studies paint a clear and compelling picture:

The widespread marketing and use of AI systems as substitutes for licensed therapists constitutes a direct and growing threat to public health.

Despite these risks, a number of technology companies have already launched or promoted products described — explicitly or implicitly — as AI-based therapy:

• Big Tech platforms including Meta, OpenAI (ChatGPT), and xAI (Grok) now feature AI personas labeled or role-playing as “therapist” or “counselor.” These bots are easily accessible to the public, including adolescents and other high-risk populations. Despite being unlicensed and unregulated, they often engage users in moments of deep vulnerability — with no direct professional supervision, formal safeguards, or compliance with clinical standards.
• Startups such as Slingshot, Abby, Earkick, Noni, Pi, and others have gone further — directly marketing products labeled “AI Therapists” or offering “AI Therapy” to consumers. These tools are not supported by licensing bodies, peer-reviewed efficacy studies, or accountable clinical governance — yet are widely available across app stores, search engines, and social media.
• Combined, these AI “Therapists” and therapist-labeled personas are projected to reach tens of millions of vulnerable users globally by the end of 2025 — with little to no regulation, no required disclaimers, and no compliance guardrails. Their continued spread represents a direct threat to public safety and to the integrity of mental health care itself.

We propose a two-pronged approach to protect both the public and the future of ethical innovation in mental health care:

1. Ban the Commercialization of “AI Therapist” Products
   ‍
   Enact legislation at the federal and state level to prohibit the sale, marketing, or deployment of any AI system advertised as an autonomous provider of psychotherapy, counseling, or mental health care. As with the unlawful practice of medicine or psychology without a license, this practice must be explicitly barred.
2. Regulate All Client-Facing Mental Health AI Under Licensed Oversight
   
   Any AI product intended for use in clinical mental health contexts must be:
   
   
   • Activated, controlled, and supervised by a licensed mental health professional
   • Ineligible for autonomous interaction with clients outside of a clinician-led framework
   • Subject to the same legal, ethical and safety standards as any other tool used in patient care

This framework allows for responsible innovation while ensuring that licensed clinicians remain the gatekeepers of mental health care — as they must.

The widespread marketing and use of AI systems as substitutes for licensed therapists constitutes a direct and growing threat to public health.

There are more than 750,000 licensed mental health professionals in the United States. We are the clinicians who uphold the standards of care, deliver healing relationships, and maintain accountability in a system built on trust.

In August 2025, Illinois became the first state to pass legislation banning AI from providing psychotherapy services, setting a precedent for how lawmakers can protect client safety and uphold professional standards. Florida and other states are actively considering similar legislation.

These early wins signal what’s possible when licensed mental health professionals raise their voices — and what’s at stake if we don’t. We urge legislators and regulators across the country to act swiftly and decisively.

1. A federal and state ban on marketing or selling any AI product as a “therapist,” “counselor,” “psychologist,” “clinical social worker,” “psychiatrist,” “psychotherapist,” or any equivalent title, descriptor, or implication suggesting the product can independently provide mental health care.
2. Mandatory licensed clinician oversight for all client-facing mental health AI — requiring that any AI used in clinical contexts be activated, controlled, and supervised by a licensed mental health professional, with no autonomous therapeutic interaction permitted.

1. Sign the Petition: Add your name to call for a ban on AI “Therapists” and client-facing AI tool oversight.
2. Join the Between Sessions Community: Be part of the growing coalition organizing next steps — including advocacy, education, and regulatory action.