Utah Is Letting AI Prescribe Psychiatric Meds. And I Have Thoughts.
So apparently we're here now. Utah just approved an AI to renew psychiatric prescriptions — no doctor needed. First time this has happened anywhere in the world.
The startup behind it is called Legion Health. They charge $19/month. Their AI can refill 15 mental health medications — things like Prozac, Zoloft, Wellbutrin — for patients who are already stable and already have an existing prescription from a human psychiatrist. No new prescriptions. No dose changes. No controlled substances.
Honestly? On paper it sounds fine. There's a real shortage of psychiatrists in Utah — around 500,000 people live in areas with no proper mental health access. If someone just needs their monthly Sertraline renewed and they've been stable for years, does that really need a doctor's time? Maybe not.
But here's what makes me a little uneasy.
This is actually Utah's second AI prescribing experiment. The first one, a startup called Doctronic, launched in January for things like blood pressure meds and statins. Within weeks, security researchers managed to jailbreak it — got it to triple an OxyContin dose, recommend meth as a treatment, and push vaccine conspiracy theories. Not great.
Utah says the production version was safer. The researchers said that's not really the point — LLMs can always be manipulated, you're just making it harder.
And now we're doing the same thing but with psychiatric medications.
To be fair, Legion has stricter guardrails. The first 250 prescriptions get reviewed by a physician before anything reaches a pharmacy. Monthly reports to regulators. Automatic handoff to humans if the AI detects suicidal thoughts, mania, or severe side effects.
But a psychiatrist from the University of Utah made a point that stuck with me — the people this system can help are the easy, stable cases that most doctors would already handle quickly anyway. The people who actually struggle to get care? They're excluded by the eligibility rules.
So who is this really for?
I don't think AI in healthcare is inherently bad. Used as a tool alongside doctors, it can do a lot of good. But autonomous prescribing — even narrow, even monitored — is a different thing. And we're learning in real time whether that's okay.
Utah's pilot runs for 12 months. Let's see how it goes.
Sources: