Mental healthcare is becoming the proving ground for AI regulation

Starting on August 12, mental healthcare providers in Colorado will be subject to new limitations around the use of AI as part of their therapeutic and clinical services.  

HB-26-1195 states that regulated psychotherapy professionals: 

“…Shall not use an AI system to provide, direct, or guide psychotherapy, clinical intervention, counseling, diagnosis, treatment planning, or other activity that constitutes the practice of psychotherapy with an individual or group,” and will be prohibited from using an AI system to interact with clients “…in any form of therapeutic communication without synchronous, real-time interaction between the regulated professional, the AI system, and the client, or generate therapeutic recommendations or treatment plans without review and approval by the regulated professional.”   

It’s one of several state laws coming into effect this summer designed to put explicit legal parameters around the use of AI in mental healthcare, and comes at a time when the rapid rise of AI for mental and emotional care and support is changing the conversation around the nation’s ongoing mental healthcare crisis. 

Filling unmet demand or preempting the human-to-human relationship?

Mental healthcare represents one of the most ethically fraught use cases for AI, not least because it has quickly become one of the most attractive use cases for consumers. 

About 16% of people have used a consumer-grade AI tool to seek information about their mental health, including more than a quarter of people aged 18-29. 

The use of AI chatbots for mental health care among adolescents and young adults in particular has increased 60 percent in just a year, according to a new survey from Harvard Medical School, translating to about one-in-five people aged 12 to 21. 

Some experts believe that AI chatbots can help fill the widening gulf between the need for mental health support and the availability of qualified human professionals. For example, a study out of Dartmouth in 2025 found a fully generative AI therapy system helped significantly reduce symptoms of depression and anxiety as well as eating disorder risks. 

With 47% of Americans living in designated mental healthcare workforce shortage areas, it’s not surprising that developers and health systems are eager to explore how similar AI tools can expand capacity for millions of people in need. 

However, there is also deep concern about how to safely and effectively close those gaps – especially since many users are engaging with generalized consumer tools that are not specifically designed to provide appropriate support.  

This may be especially worrying for youth who are often using these tools without oversight, said Hao Yu, Harvard Medical School associate professor of population medicine at the Harvard Pilgrim Healthcare Institute and leader of the Harvard survey. 

“It may surprise some that we found the majority of youth using AI did not disclose their use to any adults, either parents or physicians,” he noted. “That might be concerning, because there is misinformation available through different AI models, which are not designed to provide mental health advice,”  

Some openly available AI chatbots have already failed key measures of safety and have been shown to exhibit bias against people with mental health conditions and miss critical warning signs of crisis.  

Meanwhile, there is a growing number of headlines about self-harm and suicide allegedly tied to AI use, as well as warning rumbles from mental healthcare providers themselves about the risks of allowing AI to co-opt their clinical judgement and patient relationships.  

Rapid fire legislative action to structure the AI mental health ecosystem

With a growing need for oversight and accountability – and a lack of definitive action from federal authorities as the Administration attempts to muscle through its own AI agenda – state legislatures aren’t waiting around to start laying down the rules of the road. 

Instead, they are taking action to define the parameters of the mental health profession and codify AI’s place within it.   

A number of state laws are going into effect in the summer of 2026 and into early 2027. The trend appears to be positioning AI as a useful administrative and communication tool while clearly limiting its ability to take on any of the diagnostic and interpersonal support functions of a human clinician without full and active engagement and monitoring from clinicians themselves. 

“AI chatbots are being mistaken by patients as legitimate therapy, which is why we’re stepping up to ensure Coloradans are protected and informed,” said Colorado State Rep. Gretchen Rydin (D-Littleton). “Going into effect this month, HB26-1195 establishes reasonable protective measures on AI use in mental and behavioral health care, including prohibiting the use of AI to independently provide therapy to a patient. This law does not hamper licensed professionals’ use of AI for administrative purposes, but does require all clinical treatment and psychotherapy to be administered by a licensed, human provider.” 

In addition to Colorado’s law, Vermont enacted legislation in June 2026 that “prohibits a corporation or entity from providing, advertising, other otherwise offering mental health services, including through the use of artificial intelligence, to the public unless the mental health services are either provided by a mental health professional or part of an approved institutional review board or privacy board study.”   

The law adds that mental health professionals are permitted to use HIPAA-compliant AI tools “if the mental health professional first reviews and approves any mental health services,” which leaves sufficient room for providers to creatively incorporate specific AI tools into their practice. 

Rhode Island will enact a similar law in January of 2027, and has paired that legislation with a law that will require a chatbot operator to include protocols for addressing possible suicidal ideation or other indications of self-harm or harm to others expressed by a user. 

“As soon as any of those expressions are made, the chatbot operator must provide a notification to the user that refers them to crisis service providers such as a suicide hotline, crisis text line, or other appropriate crisis services,” explains a report by The Transparency Coalition. “Operators must file annual reports with the state attorney general that includes the number of safety protocol activations.” 

Tennessee’s take on the issue “prohibits a person from developing or deploying an artificial intelligence system that advertises or represents to the public that such system is or is able to act as a qualified mental health professional,” and complements the law with a clear definition of personhood for statutory reasons that “excludes artificial intelligence, a computer algorithm, a software program, computer hardware, or any type of machine from the definition.” 

Several other states are considering legislation around this issue, as well, including Indiana, Alabama, Missouri, Florida, and Vermont.  Most are designed to place some type of restriction on AI chatbots acting as therapists or mental health providers in an effort to preserve the human role in this type of care. 

A fragmented environment, albeit with largely shared goals

For developers and health systems, there is an inherent challenge in this legislation happening at the state level. Tools that are intended to be rolled out to a national customer base will need to meet each state’s individual parameters for operations, which could get tricky as more and more states enact their own slightly different versions of similar laws. 

The complexity is compounded by the fact that federal oversight for AI tools is split among multiple agencies and constantly evolving, with many types of emerging systems falling through existing cracks in regulatory definitions. 

The fact is that the technology – and its user base – are outpacing legislative action and policy consensus, even if state lawmakers are generally aligned on their views about how AI should be slotted into the mental healthcare environment.  

Mental healthcare is becoming one of the real tests of how legislation can address unintended consequences of AI at the population level. The policy frameworks taking shape today will almost certainly influence how lawmakers approach AI in other areas of healthcare and beyond, making these early efforts an important proving ground for balancing innovation with safety, transparency, and accountability. 

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at [email protected].