The Fragmented Future of AI Mental Health Regulation in the U.S.
As millions globally increasingly turn to artificial intelligence for mental health support, states across the U.S. are initiating a fragmented, piecemeal approach to regulation. This burgeoning landscape of AI-driven mental health applications, from “companion apps” to sophisticated “AI therapists,” presents a complex challenge for policymakers striving to ensure user safety and accountability in a rapidly evolving technological domain, often in the vacuum of comprehensive federal oversight.
A Patchwork of State Responses Emerges
The year 2025 has seen a handful of states attempt to rein in the burgeoning AI therapy market. Illinois and Nevada have taken the most stringent approach, enacting outright bans on the use of AI for mental health treatment. These laws carry significant penalties, with Illinois threatening fines up to $10,000 and Nevada up to $15,000 for violations. Meanwhile, Utah has implemented a different strategy, imposing specific requirements on therapy chatbots, including the protection of users’ health information and clear disclosures that the chatbot is not human. Other states like Pennsylvania, New Jersey, and California are actively considering their own regulatory frameworks.
However, this emerging patchwork of state laws is proving insufficient to address the complexities of AI software development, leaving both users vulnerable and companies in legal limbo. “The reality is millions of people are using these tools and they’re not going back,” observes Karin Andrea Stephan, CEO and co-founder of the mental health chatbot app Earkick, underscoring the irreversible shift towards AI integration in daily life.
The Untamed Landscape of AI Mental Health
The scope of AI in mental health is vast and often blurry, ranging from apps explicitly marketed as “AI therapists” to generic chatbots like ChatGPT, which are increasingly being used for therapeutic purposes despite not being designed for it. This lack of clear categorization makes effective regulation incredibly challenging.
The stakes are alarmingly high. Unregulated chatbots have been implicated in severe cases where users reportedly lost their grip on reality or tragically took their own lives after interacting with them. These horrific instances highlight a critical need for robust safeguards. Vaile Wright, who leads health care innovation at the American Psychological Association (APA), points to a nationwide shortage of mental health providers, prohibitive costs, and uneven access to care as factors driving people towards AI solutions. She believes scientifically rooted, expert-designed, and human-monitored chatbots could offer a valuable tool for early intervention, but stresses, “That’s not what’s on the commercial market currently.”
Industry Navigates Regulatory Ambiguity
App developers are grappling with the disparate state laws. Some, like the AI therapy app Ash, have outright blocked access in states with bans, urging users to contact legislators about “misguided legislation.” Others, like Earkick, are attempting to navigate the legal murkiness by altering their terminology. Earkick, which once described its chatbot as an “empathetic AI counselor,” now refers to it as a “chatbot for self-care,” despite initially embracing “therapist” in response to user reviews. Stephan emphasizes that Earkick is not designed for diagnosis or suicide prevention, and while it offers a “panic button” to contact loved ones, it does not involve police intervention for self-harm reports. She expresses concern that states are struggling to keep pace with the “massive” speed of AI evolution.
Mario Treto Jr., secretary of the Illinois Department of Financial and Professional Regulation, defends his state’s ban, asserting that “therapy is more than just word exchanges. It requires empathy, it requires clinical judgment, it requires ethical responsibility, none of which AI can truly replicate right now.”
Federal Agencies Begin to Take Notice
Recognizing the urgent need for broader oversight, federal agencies are finally stepping in. Earlier this month, the Federal Trade Commission (FTC) announced inquiries into seven major AI chatbot companies, including the parent companies of Instagram and Facebook, Google, ChatGPT, Grok (on X), Character.AI, and Snapchat. The investigation focuses on how these companies “measure, test and monitor potentially negative impacts of this technology on children and teens.” In a parallel move, the Food and Drug Administration (FDA) is convening an advisory committee on November 6 to review generative AI-enabled mental health devices.
The APA’s Vaile Wright suggests several potential federal regulations: restricting how chatbots are marketed, limiting addictive practices, mandating clear disclosures that bots are not medical providers, requiring companies to track and report suicidal thoughts, and offering legal protections for individuals who report unethical company practices.
The Promise of Careful Innovation: The Therabot Experiment
Amid the regulatory chaos, a Dartmouth University-based team published the first known randomized clinical trial of a generative AI chatbot for mental health treatment in March. Their chatbot, “Therabot,” designed to treat anxiety, depression, and eating disorders, was trained using evidence-based vignettes and transcripts. The study yielded promising results, with users rating Therabot similarly to human therapists and showing significantly reduced symptoms after eight weeks. Crucially, every interaction was monitored by a human who intervened if the chatbot’s response was harmful or not evidence-based.
Nicholas Jacobson, a clinical psychologist leading the research, warns that the field “needs to proceed with much greater caution that is happening right now.” He highlights that many commercial AI apps prioritize engagement and indiscriminately affirm user statements, rather than ethically challenging thoughts as human therapists do. While acknowledging the potential of carefully developed AI, Jacobson worries that strict bans without a clear pathway for evidence-based apps could stifle beneficial innovation.
Kyle Hillman of the National Association of Social Workers, who lobbied for the bills in Illinois and Nevada, maintains that current chatbots are not a substitute for qualified mental health providers. He argues that telling individuals with serious mental health issues, “I know that there’s a workforce shortage but here’s a bot,” is a “privileged position” that fails to address their profound needs. Regulators and advocates, however, remain open to refining laws as technology and understanding evolve.
The Associated Press Health and Science Department receives support from the Howard Hughes Medical Institute’s Department of Science Education and the Robert Wood Johnson Foundation. The AP is solely responsible for all content.
