Four states now ban AI therapy. Thirty-six are writing rules.
The mental-health app category spent a decade asking to be treated as software. In 2026 it started being treated as care — state by state, with very different rules in each.
As of July 2026, four states — Illinois, Nevada, Rhode Island and Maine — prohibit AI from delivering therapy. California's law, effective 1 January 2026, requires operators to screen for suicidal ideation and allows users who allege harm to sue. In the first quarter of 2026, 36 states introduced more than 70 bills regulating AI chatbots, most requiring disclosure that the user is talking to a machine.
The facts, in brief
| States banning AI-delivered therapy | Illinois, Nevada, Rhode Island, Maine |
| Illinois statute | HB 1806 — AI therapy only via a licensed professional, with written notice and consent |
| California | Effective 1 Jan 2026 — suicidal-ideation screening required; private right of action |
| Utah | First mover, May 2025 — disclosure-first, no ban |
| Q1 2026 legislative activity | 36 states, 70+ bills |
| Federal | FTC 6(b) orders issued Sept 2025 to seven companies on companion chatbots and minors |
What do the laws actually require?
They diverge sharply, which is the compliance problem. Illinois HB 1806 restricts AI for therapy or psychotherapy unless it is delivered by a licensed professional who informs the patient in writing and obtains consent, and it bars licensed professionals from letting AI make independent therapeutic decisions or interact directly with clients — AI is permitted for administrative or supplemental work under supervision.
Utah went first, in May 2025, and banned nothing. Its framework is disclosure-led: a mental-health chatbot must tell you clearly that it is AI at first contact, again when you return after a week away, and any time you ask. California's approach is different again — an operator must screen for suicidal ideation, and a user alleging harm can sue.
Ban, disclose, or screen-and-be-liable: three different theories of the same risk, now live in the same country.
Why now?
Because companion chatbots got good enough to be mistaken for care, and because the federal picture moved first. In September 2025 the FTC issued 6(b) study orders to seven companies — including OpenAI, Meta, Alphabet and Character Technologies — demanding information on how companion chatbots affect children and teens and what safeguards exist. States did not wait for the answers.
What it means if you use a mental-health app
Two practical things. First, the disclosure you now see at the start of a chat is a legal artefact, not a design choice — and its presence tells you the operator is paying attention to the rules. Second, an app's behaviour may legitimately differ depending on the state you're in, which is new and largely undisclosed.
None of this settles whether these tools work. That's a separate question, and one where the published evidence remains thin — as we found comparing Headspace and Calm, where a systematic review located 14 randomised trials for one app and a single trial for the other.
bills regulating AI chatbots introduced across 36 states in the first quarter of 2026 alone. Most ask for the same thing: tell the user they're talking to a machine.State legislative trackers, 2026
Applandica's read
Hannah Weiss · Reviews EditorWhat strikes me about this wave is how modest most of it is. Strip out the four bans and the dominant demand is disclosure — say you're a machine, say it again later, say it when asked. That such a low bar required seventy bills tells you how the category behaved when nobody was asking.
My worry is the opposite of the industry's. A patchwork this uneven encourages apps to build for the loosest state and geofence the rest, which means the protection you get depends on your billing address. I'd rather see one clear federal disclosure standard than fifty careful ones.
Frequently asked questions
Which states ban AI therapy?
As of July 2026, Illinois, Nevada, Rhode Island and Maine prohibit AI from delivering therapy. At least six states have taken action of some kind against chatbots delivering something close to mental-health care.
What does the Illinois AI therapy law say?
Illinois HB 1806 restricts AI use for therapy or psychotherapy unless delivered by a licensed professional who provides written notice and obtains consent. It prohibits AI from making independent therapeutic decisions or interacting directly with clients, allowing it only for administrative or supplemental tasks under professional oversight.
Are meditation apps like Calm and Headspace affected?
Generally no — these laws target chatbots that deliver something resembling therapy or psychotherapy, not guided-meditation libraries. Apps adding conversational AI coaching should read the statutes carefully.
Sources
- Quartz — State laws restricting AI in mental health care, explained. Accessed 1 Sep 2026
- Bloomberg Law — AI therapy chatbots spur states to act over patient safety fears. Accessed 1 Sep 2026
- Psychology.com — Illinois banned AI therapy: what the law actually means. Accessed 1 Sep 2026
FTC sues Hims & Hers over health data sent to Meta and Snap
The complaint pairs the two failures we log most often — advertising trackers on sensitive health data, and a…
AnalysisFDA steps back from wearables and AI health software
A January guidance shift means many wearables and AI health tools can reach the market without formal device review.…