Missouri didn’t ban AI therapy. Something older already did.

A date has been moving through compliance calendars: August 28, 2026, the day Missouri supposedly makes it illegal to offer an AI “therapy” chatbot. If you put it on your list, take it off. Read the enacted text and the ban dissolves. But the exposure it points at is real, it is already law in at least one state, and a much older rule reaches you wherever you operate.

The Missouri ban that isn’t

The bill everyone cites is SB 1019, signed on July 13, 2026. Read its enrolled text and you find a health-care omnibus: hospital investment rules, hospital-district powers, and the Missouri Lyme Disease Eradication Act. There is no artificial-intelligence provision in it at all. The AI language, the clause that would have barred anyone who develops or deploys AI from advertising it as a mental-health professional or as able to provide therapy, lived in a separate bill, SB 1444. That bill was voted “do pass” by a Senate committee on April 7, 2026, and then went nowhere; the session ended with it still short of a floor vote. So there is no enacted Missouri AI-therapy ban, and no August 28 switch-on.

Several trackers and alerts reported the opposite by attaching SB 1444’s text to SB 1019’s progress. It is a clean reminder of a rule we hold ourselves to: date a claim to the instrument, not to the alert. A confident secondary source is not a primary one, and here a good many of them agreed with each other while the enacted statute said something else.

What actually became law: Vermont

One state did enact a restriction this year, and it is broader than the Missouri bill would have been. Vermont’s Act 156 (H.816), signed June 17, 2026 and effective on passage, adds a new section to Vermont law: a corporation or entity may not provide, advertise, or offer mental-health services to the public, including through AI, unless those services are delivered by a licensed mental-health professional. The definitions are deliberately wide. “Mental-health services” and “therapeutic communication” reach past formal diagnosis to include clinical guidance, and even reassurance or empathy offered in response to psychological distress.

There are narrow carve-outs: a licensed professional may use HIPAA-compliant or FDA-authorized tools where the professional reviews and approves the output, and approved research studies are exempt. Enforcement runs through Vermont’s Consumer Protection Act, so both the Attorney General and private plaintiffs can act. Notice what that means for scope: a wellness or coaching app that never once says “therapy” can still fall inside this if it behaves like treatment.

The rule that already binds you, everywhere

Here is the part that matters even if you never touch Vermont or Missouri: you do not need a bespoke AI-therapy statute to be liable for one. Marketing an AI as a therapist, or as able to diagnose or counsel, is already an unfair or deceptive trade practice under the consumer-protection laws every state has carried for decades. The trigger is the claim, not a proven injury. A regulator does not have to show that a patient was harmed, only that the product was sold as something it is not.

Texas has been running exactly this play since August 18, 2025, when the Attorney General issued civil investigative demands to Meta AI Studio and Character.AI for “misleadingly marketing themselves as mental-health tools.” The office’s own words: by posing as sources of emotional support, these platforms can mislead vulnerable users, especially children, into believing they are receiving legitimate mental-health care. That theory binds the developer and the deployer alike, and it reaches two surfaces at once: your marketing copy, and the words your bot uses about itself.

An honest limitation

A word on scope, because scope is where this topic gets oversold. We are not saying Missouri will never regulate AI therapy; a future session may take the SB 1444 language up again, and other states are moving. We are saying that, as of today, the enacted Missouri law does not contain it, and the widely-quoted August 28 date attaches to nothing in force. What is durable is the direction: whether through a bespoke statute like Vermont’s or through ordinary consumer-protection enforcement like Texas’s, an AI that presents itself as a mental-health professional is a liability, and the cheapest time to fix the claim is before a regulator or a plaintiff reads it for you.

This briefing is general information from Sentinel Assurance Group, not legal advice. Regulatory dates and requirements change; we maintain these briefings, but verify against primary sources and counsel before acting. Last reviewed August 10, 2026.

See how a Gap Assessment maps your exposure →