Новый закон Колорадо требует от терапевтов постоянного контроля за использованием ИИ, что может парализовать профессию

In today’s column, I examine a new law enacted by Colorado that requires therapists who use AI with their clients as part of their psychotherapy to be immersed in the client-AI interactions at all times. The law says that a therapist must be actively logged into the AI and paying attention on a real-time basis, ready to intervene instantly whenever needed. Though this might seem at first glance to be an abundantly helpful safeguard, there are several significant, perhaps unintentionally adverse consequences that don’t seem to have been weighed suitably. First, this real-time requirement is going to be a tremendous time-sink burden on the therapist and raises difficult questions of whether their time is being usefully undertaken, and whether it will be reimbursed by standard mental health coverages. Second, this brazen law will discourage therapists from making use of AI for their therapeutic efforts, and they will undoubtedly avoid incorporating AI into their professional efforts (stagnating the profession). Third, clients will likely opt to use AI entirely on their own, rather than having to pay their therapist during the oversight AI time, and thus the law is ironically pushing clients toward unfettered AI instead of toward AI that is carefully handpicked and generally monitored by the therapist. It’s a mess. This recently passed law is yet another example of my ongoing exhortation that new AI laws are being drafted and enacted without sufficient attention to what kinds of results they may produce. I would estimate that the hearts of policymakers and lawmakers are typically in the right place, but they do not perform deep enough due diligence to ferret out the downsides and fix potential legal holes before turning bills into actual hard-nosed laws. Let’s talk about it. This analysis of AI breakthroughs is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here). AI And Mental Well-Being As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the advent of modern-era AI that produces mental health advice and performs AI-driven therapy. This rising use of AI has principally been spurred by the evolving advances and widespread adoption of generative AI. For an extensive listing of my well over one hundred analyses and postings, see the link here and the link here. There is little doubt that this is a rapidly developing field and that there are tremendous upsides to be had, but at the same time, regrettably, hidden risks and outright gotchas come into these endeavors, too. I frequently speak up about these pressing matters, including in an appearance on an episode of CBS’s 60 Minutes; see the link here. AI Providing Mental Health Guidance Millions upon millions of people are using generative AI as their ongoing advisor on mental health considerations (note that ChatGPT alone has over 900 million weekly active users, a notable proportion of whom dip into mental health aspects; see my analysis at the link here). The top-ranked use of contemporary generative AI and LLMs is to consult with the AI on mental health facets; see my coverage at the link here. This popular usage makes abundant sense. You can access most of the major generative AI systems for nearly free or at a super low cost, doing so anywhere and at any time. Thus, if you have any mental health qualms that you want to chat about, all you need to do is log in to AI and proceed forthwith on a 24/7 basis. There are significant worries that AI can readily go off the rails or otherwise dispense unsuitable or even egregiously inappropriate mental health advice. Banner headlines last year accompanied the lawsuit filed against OpenAI for their lack of AI safeguards when it came to providing cognitive advisement. Today’s generic LLMs, such as ChatGPT, GPT-5, Claude, Gemini, Grok, CoPilot, and others (all known as general-purpose AI or GPAI), are not at all akin to the robust capabilities of human therapists. Meanwhile, specialized LLMs are being built to attain similar qualities (known as purpose-built AI or PBAI), but they are still primarily in the development and testing stages. See my coverage at the link here. Various State Laws On AI Mental Health A beehive of activity is taking place regarding crafting new AI laws. See my extensive coverage of state-level AI mental health laws at the link here. It is a matter on the minds of the public and in the hands of the state legislators. Some people ardently believe that AI and AI makers are being allowed to run amok. New AI laws are vitally needed to protect society from this onslaught of ubiquitous AI. I previously examined a notable AI and mental health law passed last year by Illinois see the link here, one that was also enacted by Nevada see the link here, and one that was enacted by Utah see the link here. Those laws are scoped to prevail within their respective state boundaries. In that sense, these laws are applicable to AI usage within the particular state and do not bear on other states per se. Big Picture Of AI Mental Health Laws Not everyone agrees with this pell-mell rush of new AI laws, or at least they are concerned that these AI laws might go overboard. In the zeal to protect society, there is a chance that we might unduly restrict innovation and delay or undercut the benefits of leading-edge AI. The debate is ongoing and heated. Readers might recall that I proposed a 7-step AI-law-making process that I believe could substantively help regulators to devise new AI laws that are on target and balanced; see my depiction at the link here. This has the added benefit of reducing what I refer to as AI-law legal debt. This refers to AI laws that, though they look shiny, contain hidden debt that must ultimately be paid. Legal glitches and hitches will eventually be found when AI laws are passed without suitable scrutiny and analysis. My prediction is that the slew of newly passed AI laws is likely to create a legal quagmire in the courts. In terms of the AI laws in the United States, they have not yet stood the test of time, meaning that we won’t really know how well they stand up until there are court cases that test these new laws. It is too early to know whether the laws will survive legal battles waged by AI makers and other contenders. Just because AI laws are enacted does not mean they are proper. All sorts of improper provisions and constitutionally contentious stipulations are undoubtedly buried within these shiny new AI laws. Congress has repeatedly waded into establishing an overarching federal law that would encompass AI. So far, no dice. The efforts have ultimately faded from view. Thus, at this time, there isn’t an overarching federal law devoted to these controversial AI matters. The big question will be to what degree a sweeping federal law would impact the numerous state-level AI laws. The odds are that many of the state-level laws would run afoul of a federal mandate, and a tsunami of legal cases would arise as a tussle between federal law and state law is undertaken. It surely will be a legal mess. Colorado Passes AI Mental Health Law Colorado recently passed House Bill 26-1195, known as HB1195, and the legislation was signed into law on June 3, 2026. This newly enacted legislation is known as Section 12-245-224.5 and is formally entitled “Use of artificial intelligence systems for psychotherapy services – permitted use by regulated individuals – consent – use in educational and training settings and research – prohibited use – definitions.” Much of the new law is relatively typical of other AI-related mental health laws at the state level. I won’t go into the portions that are of a usual nature. You are encouraged to consider reading the full bill if you are generally interested in these types of laws. My focus goes to the segment that does something extraordinary and isn’t yet commonly undertaken, though I am certainly concerned that it might catch on and be used or copied into other state-level laws. The passage in question is item 5, and especially part 5a: - “(5) A licensee, registrant, certificate holder, or other individual lawfully permitted to provide psychotherapy services in the state shall not allow any artificial intelligence system to do any of the following:” - “(a) Interact with clients in any form of therapeutic communication without synchronous, real-time interaction between the licensee, registrant, certificate holder, or other individual lawfully permitted to provide psychotherapy services in the state, the artificial intelligence system, and the client; or” - “(b) Generate therapeutic recommendations or treatment plans without review and approval by the licensee, registrant, certificate holder, or other individual lawfully permitted to provide psychotherapy services in the state.” Before we unpack 5a, allow me to first note that 5b is somewhat routine or regularly stipulated. Other state-level AI-related mental health laws will often require that if AI is used as a therapeutic tool with clients, the therapist must first review and approve any treatment plans or therapeutic recommendations before a client is given or told to use the AI-generated approach. It makes clear-cut sense that the therapist should be deciding whether the AI has satisfactorily provided appropriate recommendations. Unpacking The Item 5a Part 5a is quite a head-turner. As you can see, the stipulation is that if a therapist opts to use AI as a client-facing therapeutic tool, the therapist must co-participate in the AI tool usage with the client on a synchronous, real-time, interactive basis. This type of requirement is highly unusual. Let’s consider what it is intended to accomplish. The logical case for the stipulation is that it is prudent as a means of protecting a client from potential mental harm or cognitive disturbance caused by what the AI says during a chat. The therapist is going to be right there online, watching the chat, and ready to step in. If the AI goes off the rails, the therapist will presumably stop the chat and inform the client that they are to ignore what the AI had to say. In addition, the therapist seemingly would aid in steering the AI during a chat, trying to keep the AI aimed in a fashion matching whatever treatment path the therapist deems warranted. So, yes, the idea of forcing a therapist to be in the loop during an AI chat that their client is having has substantive merit. The idea makes apparent sense. But there are downsides that need to be surfaced and compared as a tradeoff to this considered safety provision. Trust Is On The Line One perspective is that the 5a portion of the law is shaped around a distrust of what AI might do. That’s fine. We all can agree that AI will, at times, encounter so-called AI hallucinations, which are confabulations that could produce fake or false advice for a client. Generative AI is also known to have difficulties in lengthy chats and will tend to drift from guardrails. There is plenty of basis for distrusting what AI might do during a mental health chat. The unstated aspect is that 5a seems to be saying that the therapist isn’t trustworthy either. How so? If a therapist has dutifully selected AI as a client-facing tool, they will have presumably done their due diligence and decided that they can allow the AI to work directly with the client and that only particular circumstances might require their active involvement. This is a therapy consideration by the therapist. The key is that the therapist is where the buck stops. It should be on the shoulders of the therapist to decide what is safest and best for their client in this regard. Instead, the law is essentially asserting that a therapist cannot be trusted to properly make use of AI as a client-facing tool. Therefore, the law will step in and force therapists to be online and active whenever they have a client making use of AI for mental health purposes. Use Of Therapist Time Is this even a suitable use of a therapist’s time? The question is important since a therapist having to be logged into the AI and attentive on a real-time, synchronous basis will really have no choice but to devote their time to that task. The law seems to suggest that the therapist cannot be doing double-duty and working on other tasks, i.e., they presumably are to be riveted to what the AI chat is saying. A well-designed AI for mental health would readily alert a therapist if something is amiss or provide snippets along the way about how things are going. A therapist would not need to be fully devoted to being avidly in the loop. The gist is that under this law, the therapist is going to be using their time solely to be in the AI chat as a monitoring mechanism. You could contend that this is underuse or misuse of the therapist’s time. Another related aspect is whether a therapist can be compensated for this use of their time in this particular manner. You might claim that since the therapist is seemingly not performing the therapy, which the AI is perhaps doing so, and is merely being a monitoring observer, perhaps the therapist should not be paid for this. Or be paid some lesser amount in contrast to their regular human-to-human therapy session rate. Liability Exposure Widens This also opens an additional liability exposure for therapists. Imagine this. A client is using AI, and the therapist is in the loop. Days later, the client believes that the AI chat told them something that it shouldn’t have said. The client insists that the therapist failed to intervene on a real-time synchronous basis as is required by this new law. Bam, the therapist is now trapped by the law. The potential malpractice inquiry might go this way. Was the therapist paying attention? Why didn’t they intervene? It was the legal duty of the therapist to do so. In contrast, if this part 5a didn’t exist, the exposure would seemingly not be as heightened and readily triggered. AI Becomes A Danger For The Therapist All told, this part 5a is going to heavily discourage therapists from using AI as a client-facing tool. The risks versus the rewards are not worth the effort. I would wager that most therapists who are considering using AI as a therapeutic tool will think twice about whether to go the AI route. It makes little practical sense to take such a chance. For many therapists who are already avoiding AI or don’t want to deal with it, this new law provides ideal cover for them. They can say that they would love to use AI, but the onerous provisions make it untenable. This lets those foot-dragging therapists off the hook, and they can continue to keep their heads in the sand about the possible use of AI in their therapeutic process. I’ve stated that the world of therapy is moving away from the classic dyad of therapist-client and transforming toward a new triad known as therapist-AI-client; see my in-depth analysis at the link here. The future is that AI will be intertwined in the therapeutic process, and therapists will demonstrably use AI as an integral element of their services. Part 5a is regrettably going to put a dent in that progress. Clients Going To Use AI Anyway Perhaps the worst adverse consequence is that clients will likely opt to secretly use AI anyway. Here’s the problem. Right now, people often use AI for mental health advice and then seek out a therapist after realizing that perhaps they need professional assistance. I’ve emphasized that therapists should right away find out if a prospective client is or has been using AI for their mental health guidance. If so, the therapist should take that AI usage into the therapy plan, such as considering what the AI already told the client about their mental health status. A big distinction is that if someone has been using general-purpose AI for their mental health advice, this is far different than using a purpose-built AI for mental health. A therapist might have vetted and decided upon using a PBAI that they believe is well-devised and will be helpful to the client. Part 5a seems to make no distinction between a GPAI and a PBAI. All AI seems to be the same in the eyes of this law. That’s an unfortunate flaw. In any case, with this law on the books, and assuming that therapists won’t want to have to deal with part 5a, they will simply tell the client that there isn’t any AI that is going to be used throughout their therapy. A client might be quite dismayed by this. Many will likely decide to use AI on their own. And they won’t tell the therapist about the AI usage. This part 5a is going to potentially push people toward using AI on their own, and they will keep it a secret from their therapist. This is problematic. The advice from the AI is bound to be contrary to or different from what the therapist is advising the client. Meanwhile, since the client doesn’t want to divulge the AI usage, a barrier between the therapist and the client is being formulated. It’s all a bad situation. Legally Granted Escape Hatch Therapists are between a rock and a hard place due to part 5a. But this unusual provision provides an escape hatch later in the passage that is somewhat of a wink-wink. I’m sure that therapists eager to use AI in their therapeutic process will consider using the loopholes granted in the law (this can be dicey; therefore, therapists should consult with their legal counsel accordingly). This part 9 is essentially a way around the 5a, though that’s probably not what the intention was: - “(9) This section does not apply to a licensee, registrant, certificate holder, or other individual lawfully permitted to provide psychotherapy services in the state who uses or recommends the use of a technology or service that:” - “(a) Provides self-help, therapeutic homework, patient navigation, coaching, guided meditation, journaling, reflections, psychoeducation, goal setting, progress tracking, therapeutic session preparation, therapeutic session summaries, mood monitoring, mindfulness exercises, breathing exercises, crisis resource directories, safety planning, or other wellness tools that:” - “(I) Do not diagnose or treat mental health disorders; and” - (II) Clearly and conspicuously disclose that the technology or service is not a substitute for clinical care.” The crux seems to be that if the AI is not purporting to be diagnosing or treating mental health disorders, the requirement for the therapist being actively online in real-time is no longer encapsulated (again, check with your attorney). As an aside, AI makers have wised up to the aspect that they can often get away with having their AI provide mental health advice by avoiding the wording of referring to mental health. They will say that their AI aids in mental well-being, which then often allows them to avoid the new laws restricting AI usage for mental health purposes. Some believe this is a dodge that plays on loosey-goosey semantics and should be put to a stop. Others would declare that anything goes as long as the words “mental health” are not promoted. The World We Are In The heart of the matter at hand is to protect the public from AI that can potentially give untoward or inappropriate mental health advice to users. This is a laudable goal. But tying the hands of therapists by forcing them to be logged in and acting on a real-time, synchronous basis when providing AI as a tool of therapy is not the way to proceed. Therapists should be given latitude as they professionally see fit. They are where the buck should stop. If a therapist believes that the client can access the provided AI and do so without the therapist being continuously involved, that’s on the therapist’s shoulders. They are the party responsible. New laws about AI are rapidly being passed at the state level. Policymakers and lawmakers need to closely consider what their drafts say and whether there are better ways to design and express what they intend to legally regulate. The downstream impacts can go in directions they might not have anticipated, and that, upon reflection, they realize in hindsight, were not what they hoped for. AI has a dual-use effect. Just as AI can be detrimental to mental well-being, it can also be a huge bolstering force for mental health. A delicate tradeoff must be mindfully managed. Prevent or mitigate the downsides, and meanwhile make the upsides as widely and readily available as possible. A final thought for now. Thomas Edison famously made this remark: “A good intention, with a bad approach, often leads to a poor result.” It takes a village to design, craft, and field new AI laws that are both well-intended and will produce beneficial results.