AI is moving from pilot project to daily reality across behavioral healthcare. Systems are using it to ease documentation burden, train crisis-response staff and flag patients at risk of relapse.
On the other hand, regulators, unions and clinicians are pushing back on where AI has been allowed to operate without a licensed professional in the loop, a tension that is now playing out in court filings, state legislatures and professional associations. Here’s a look at where AI is gaining traction in behavioral health, and where it’s drawing scrutiny.
Legal and regulatory flashpoints
Oakland, Calif.-based Kaiser Foundation Health Plan is facing a complaint alleging it used an autonomous digital system to triage and refer behavioral health patients in Northern California without a licensed clinician reviewing the results.
The complaint, filed with the California Department of Managed Health Care and the U.S. Department of Labor’s Employee Benefits Security Administration by the National Union of Healthcare Workers, centers on Kaiser’s “E-Visit” tool. Enrollees seeking treatment for anxiety or depression are directed to complete an online questionnaire, and NUHW alleges an algorithm assesses the responses and generates referrals and care recommendations without review by a licensed clinician, a practice the union said violates California laws governing behavioral health triage and utilization review.
The filing also alleges Kaiser is still relying on unlicensed call-center staff for behavioral health triage, tying it back to a complaint NUHW brought to the DMHC in July 2025. Beyond asking regulators to stop both practices, the union requested additional financial penalties. It noted the DMHC already fined Kaiser $50 million and required $150 million in mental health improvements over five years under a 2023 settlement.
NUHW President Sophia Mendoza framed the core issue for Becker’s as one of access — patients using the tool never get to speak with a licensed therapist who could actually evaluate their needs — and argued that’s a problem under state law no matter how the underlying technology is classified. Kaiser pushed back, saying the E-Visit tool was built with input from both therapists and NUHW representatives, functions as one option among several rather than the only path to care, and always displays a phone number so members can reach a person directly. The health system also maintains the tool isn’t used to diagnose patients, make clinical decisions or determine medical necessity — those calls, it says, stay with licensed staff.
Kaiser pointed to more than $2 billion put into mental health services, staffing and facilities since 2020, plus an AI governance framework it worked out with the Alliance of Health Care Unions that the MIT Sloan School of Management has recognized, and said it has offered NUHW a similar joint labor-management committee. The DMHC has confirmed it’s looking into the complaint. This isn’t the first time AI has put Kaiser and its unions at odds: a 2024 complaint challenged automated therapy-necessity reviews at Kaiser’s contracted network Rula Health, and in July, California Nurses Association-represented nurses protested Kaiser’s AI strategy outside the American Hospital Association’s Leadership Summit in Denver.
States move to restrict AI therapy chatbots
Regulators are not waiting for federal guidance. Five states — Colorado, Maine, Rhode Island, Tennessee and Vermont — enacted laws in 2026 restricting AI therapy chatbots, joining Illinois and Nevada, which passed similar measures in 2025.
Most of the new laws lean on states’ existing authority to license mental health professionals. Colorado’s law restricts how licensed providers can use AI in specified ways. Maine and Tennessee bar AI systems from offering or advertising themselves as therapy services. Rhode Island prohibits licensed providers from letting AI make independent treatment decisions. Lastly, Vermont bars AI from independently delivering mental health services altogether.
Clinicians are sounding similar alarms. Psychologists have pushed the Federal Trade Commission to investigate chatbots that present themselves as therapists, citing a lack of professional training built into these tools. In one instance, a teenager’s suicide was linked to an AI companion app.
Separately, an American Psychological Association survey found that more than a third of psychologists have noticed patients growing dependent on a chatbot, with a smaller share reporting patients developing distorted thinking tied to chatbot use. One study estimated that roughly 8.2 million adolescents and young adults have used an AI chatbot for mental health advice in 2025.
Day-to-day care and workforce support
Health systems report real gains from AI on the operational and workforce side. Three areas in particular stand out: easing clinician burnout, sharpening clinical decision-making and modernizing training. Here are some of the ways the sector has been talking about AI’s role with staff:
- Cleveland Clinic has adopted an AI scribe to handle note-taking during patient encounters, freeing clinicians to focus on the person in front of them. Leopoldo Pozuelo, MD, center director of adult behavioral health, called the shift “transformative” and expects AI scribe access to become a standard expectation among clinicians being recruited. A study led by Yale New Haven Health researchers and published in JAMA Network Open found that among 263 clinicians across six health systems, burnout fell from about 52% to 39% after 30 days of ambient AI scribe use, alongside nearly an hour less documentation time outside work hours each week.
- Nashville, Tenn.-based Centerstone, which operates 988 crisis lifeline services, layers an AI-supported simulation tool called ReflexAI on top of roughly 120 hours of standard crisis-staff onboarding, according to Becky Stoll, the organization’s executive vice president of crisis services. The platform lets trainers build text-based practice scenarios around emerging crisis trends — for example, a rise in a particular category of calls — so staff can rehearse before encountering similar situations live.
- Through its “No One Cares Alone” program, Renton, Wash.-based Providence created an interactive AI chatbot training tool for the system’s 125,000 caregivers. It walks employees through realistic mental health scenarios, such as noticing a colleague seems off and initiating a difficult conversation about it. The goal is to spread basic crisis-response skills beyond clinical staff to the broader workforce.
- Fort Lauderdale, Fla.-based ABA Centers built a diagnostic team and a prescriptive tool that ties therapy-hour recommendations to a patient’s individual needs. Clinicians are able to override the tool’s suggestions when their own judgment suggests otherwise.
- Norfolk, Va.-based Sentara is exploring an AI and data-science approach to spot low-risk behavioral health patients earlier in an emergency department visit, with the goal of safely discharging them home sooner, according to Ken Dunham, MD, executive director of medical services for behavioral health.
Where the field is headed
Longer-term, national investment is flowing toward making these tools safer. Providence, R.I.-based Brown University is leading a $20 million National Science Foundation-backed institute, one of five such institutes sharing a combined $100 million in NSF funding. It is aimed at developing AI assistants for mental and behavioral health settings capable of trustworthy, sensitive, context-aware interactions with individuals, at a moment when several states have already raised safety concerns about AI’s role in mental healthcare.
At the Becker's Fall Behavioral Health Summit, taking place November 4–5 in Chicago, behavioral health leaders and executives will explore strategies for expanding access to care, integrating services, addressing workforce challenges and leveraging innovation to improve outcomes across the behavioral health continuum. Apply for complimentary registration now.
