Colorado's New AI Therapy Law: Good Guardrails, Unrealistic Expectations

Artificial intelligence has arrived in mental health care. Whether therapists embrace it, ignore it, or actively discourage it, many clients are already using tools like ChatGPT and other AI assistants to better understand their symptoms, practice coping skills, journal about difficult experiences, or simply have someone—or something—to talk to during lonely moments.

Recognizing both the promise and the risks of this rapidly evolving technology, Colorado recently passed House Bill 26-1195, a new law establishing guardrails around the use of artificial intelligence in psychotherapy. The legislation makes an important statement: psychotherapy should remain a fundamentally human endeavor. AI can assist clinicians in limited ways, but it should not replace the judgment, responsibility, and therapeutic relationship provided by a licensed mental health professional.

As a licensed psychologist, I believe those protections are appropriate. Where I become more skeptical is in the law's implicit assumption that therapists have meaningful control over how clients choose to use AI outside of the therapy office.

Those are two very different issues.

The legislation appropriately limits what licensed clinicians can do. It prohibits therapists from allowing AI to independently conduct psychotherapy or generate treatment recommendations without appropriate professional oversight. These are sensible boundaries that protect the public and reinforce the importance of clinical judgment.

However, clients are increasingly making AI part of their own mental health journey, often without their therapist's knowledge.

Some use AI to organize their thoughts before therapy. Others ask AI to explain cognitive behavioral therapy concepts, mindfulness exercises, or trauma responses. Many use it as an interactive journal or brainstorming partner between sessions. I've even heard clients describe AI as a source of encouragement during difficult moments when their therapist isn't immediately available.

Whether we like it or not, this is already happening.

The challenge is that therapists have very little ability to regulate these private interactions. We cannot realistically prevent clients from consulting AI any more than we could stop them from reading psychology books, watching mental health videos on YouTube, or participating in online peer support communities.

Instead of pretending this behavior can be controlled, I believe our role should be to help clients use these tools thoughtfully and safely.

That starts with honest conversations.

If a client tells me they have been using AI to better understand anxiety, depression, trauma, or relationship problems, my first response is curiosity rather than judgment. What questions were they asking? Did the information seem helpful? Did anything feel inaccurate or emotionally unsettling? How did it influence their thinking?

These discussions create opportunities for education rather than prohibition.

AI can be remarkably useful for psychoeducation, brainstorming coping strategies, practicing communication skills, or helping someone translate confusing emotions into words. It can encourage reflection and increase engagement in treatment.

At the same time, AI has significant limitations.

It lacks genuine empathy. It cannot fully appreciate the complexity of an individual's history, culture, relationships, or lived experience. It may provide inaccurate information with unwarranted confidence. It cannot recognize subtle warning signs that an experienced clinician would notice, nor can it assume responsibility during a mental health crisis.

Perhaps most importantly, healing often occurs within the therapeutic relationship itself. Trust, emotional attunement, compassion, and the experience of feeling deeply understood remain uniquely human qualities that no current technology can replicate.

In my view, Colorado's new law gets the most important principle exactly right: psychotherapy should continue to be led by licensed professionals who remain accountable for clinical decisions.

Where the law is less convincing is in its assumption that therapists can meaningfully control how clients independently interact with AI. That ship has already sailed.

Rather than viewing AI as something to eliminate from mental health care, I believe we should help clients become informed consumers of these rapidly evolving tools. We should encourage them to bring their AI conversations into therapy, examine them together, separate helpful insights from misinformation, and integrate useful ideas into evidence-based treatment.

Technology will continue to change. The therapeutic relationship remains constant.

The future of psychotherapy is unlikely to be humans versus AI. Instead, it will be skilled clinicians helping people navigate a world in which AI is becoming another source of information, reflection, and support—while ensuring that the heart of therapy remains deeply, authentically human.