What Practitioners and Clients Should Know About AI's New Limits in Behavioral Health

The conversation around AI in mental health just got a lot more concrete. In March 2026, the White House released its National Policy Framework for Artificial Intelligence. While much of the coverage focused on broad tech policy, practitioners working in behavioral health had good reason to pay close attention. Embedded in the framework are provisions that draw a firm line around what AI can and cannot do in therapeutic contexts, and that line has real implications for how programs like TheraHive fit into the landscape of mental health support.

Where Regulators Are Drawing the Line

State-level legislation has been moving quickly. Massachusetts lawmakers are currently reviewing SB 2632, a proposal that would restrict AI from making independent therapeutic decisions in mental or behavioral health settings. Illinois enacted its Wellness and Oversight for Psychological Resources Act in 2025, which bars AI from directly interacting with clients therapeutically or generating treatment plans without licensed professional review. Nevada passed similar legislation the same year.

The federal framework signals that these state-level concerns have reached the national conversation. Rather than waiting for a patchwork of inconsistent state laws to solidify, the White House has called on Congress to develop unified standards, including specific attention to how AI tools interact with vulnerable populations in clinical contexts.

The through-line across all of these proposals is consistent: AI can assist, but it cannot replace the human clinician making therapeutic decisions. That distinction matters, and it's one that psychoeducational programs need to understand clearly.

The Difference Between Therapy and Psychoeducation

This regulatory moment is a useful one for clarifying something that often gets muddied in public conversation: the difference between psychotherapy and psychoeducation.

TheraHive is a psychoeducational DBT skills program. That means the platform teaches DBT skills through structured group learning, guided by licensed psychologists, without functioning as individual therapy or providing clinical diagnosis. Students learn skills like DBT emotional regulation techniques, DBT mindfulness exercises, and DBT distress tolerance practices in a cohort format, with the explicit goal of building real-world coping abilities they can carry into their daily lives.

That model is not what regulators are targeting. The AI restrictions taking shape in law are aimed at tools that make independent therapeutic recommendations, interact directly with clients in a clinical therapeutic capacity, or substitute for licensed clinical judgment. A skills-based group learning environment, delivered with human clinical oversight, operates in an entirely different category.

Understanding that difference is not just semantics. It shapes how programs are designed, how they're described, and how they serve people.

Why the Human Element in Group Learning Is Irreplaceable

The regulatory instinct to protect the therapeutic relationship reflects something research has consistently supported. In DBT specifically, the group skills training format is not incidental to outcomes. It is one of the mechanisms through which the skills actually work.

Research on DBT skills training has found that increased skills use fully mediated reductions in suicidal behavior, depression, and anger control in participants receiving group-based DBT. In other words, learning the skills in a structured group context, and then using them, is what drives change. A randomized controlled trial examining DBT skills training for emotion dysregulation found effect sizes of 1.86 for reductions in emotion dysregulation, with skills use statistically mediating those improvements. These are not marginal results.

What makes the group format specifically valuable is harder to reduce to an algorithm. Shared practice, peer modeling, and the felt experience of learning alongside others who are working through similar challenges all contribute to how the skills take root. AI tools can surface information and prompt reflection, but they cannot replicate what happens in a live virtual DBT skills group session.

What AI Can Reasonably Support in Psychoeducation

None of this means that AI has no place in psychoeducational programs. The distinction regulators are drawing is not between AI and no-AI; it is between AI as a clinical decision-maker and AI as a support tool under human oversight.

There is legitimate utility for AI in administrative functions, content accessibility, skill reference tools, and scheduling. AI that helps a student recall the steps of a DBT skill between sessions, or surfaces relevant psychoeducational content in response to a question, is operating as a learning resource rather than a therapist. That is meaningfully different from a system claiming to provide therapy or generating independent treatment decisions.

Research on internet-based psychoeducation has demonstrated that structured online programs can increase accessibility and patient engagement without requiring continuous therapist involvement, precisely because they function as educational tools rather than clinical interventions. The clinical value lies in the structured curriculum, the evidence-based DBT framework, and the licensed clinicians guiding the program, not in any AI layer.

What This Means If You're Considering an Online DBT Skills Program

For people exploring online DBT therapy programs or virtual DBT skills training, this regulatory moment is actually clarifying in a helpful way.

The programs worth trusting are the ones with clear human clinical leadership, transparent descriptions of what they are and are not providing, and evidence-based curriculum grounded in actual DBT research. A program run by licensed psychologists, delivering structured DBT group therapy sessions in a cohort format, with no claims of providing diagnosis or individual therapy, is not what these regulations are trying to constrain. It is what responsible psychoeducation looks like.

Research on standalone DBT group skills training has found that group skills delivery can produce meaningful clinical benefits even without the full individual therapy component of traditional DBT, including significant reductions in emotion dysregulation. That finding supports the model that structured, group-based, skills-focused psychoeducation has genuine value for people who may not have access to, or may not currently need, traditional individual DBT therapy. If you're curious about how a structured online DBT skills group works, explore our groups here.

The Bottom Line

Regulators are drawing a clear line: AI cannot substitute for licensed clinical judgment in mental and behavioral health settings. That is a reasonable line. The more interesting question for people seeking support is what human-led, evidence-based psychoeducation can offer within that boundary, and the answer is quite a lot.

DBT skills groups built around structured learning, live group delivery, and licensed clinical oversight are exactly the kind of resource the regulatory conversation is not trying to limit. They are trying to protect the space that makes those programs valuable in the first place.

{{promo-banner-1}}