AI Regulation Tracker  /  Regulation in force

Texas HB 149 in Force: Every AI-Assisted Diagnosis or Treatment Now Needs a Patient Disclosure

TRAIGA reaches well past hiring and government use. Its healthcare provision is the one in the exam room: when AI helps diagnose or treat a Texas patient, you have to tell that patient, in plain language, at or before the point of care. Here is the disclosure step to build and the boundary that separates this from the employment angle already covered.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published June 27, 2026  ·  Last updated June 27, 2026

The healthcare clause inside the headline law

TRAIGA arrived as a broad governance statute, and most of the early coverage went where the broad language went, to employers, to government agencies, and to the new oversight structure. That coverage stands, and this newsroom already carries it. This is a different provision, and it lives in a different room. Buried inside the same act is a rule aimed squarely at clinical care, and for a practicing physician it is the part of HB 149 that changes a patient encounter.

The Texas Responsible Artificial Intelligence Governance Act, passed as HB 149 in the 89th Legislature and signed on June 22, 2025, took effect January 1, 2026. Its healthcare provision is short to state and consequential to follow: when a health care provider uses an AI system in a patient's diagnosis or treatment, the provider has to tell the patient.

What you have to say, and when

The disclosure has three properties that decide whether it counts. First, timing. It has to reach the patient at or before the point of care, before the AI-assisted interaction rather than in a summary afterward. The one carve-out is an emergency, where care comes first and the disclosure follows as soon as it reasonably can. Second, clarity. It has to be in plain language, clear and conspicuous, and it cannot be buried in a design that obscures it. Third, scope. It is tied to AI used in diagnosis or treatment, which is a wider net than it first sounds. A diagnostic imaging assistant, a triage or risk-scoring tool, a clinical decision support layer, and a treatment-planning system can all fall inside it.

What the law does not require is a tutorial. The point is not to explain the model. The point is to tell the patient, before the moment of care, that AI is part of how their diagnosis or treatment is being reached.

Why this is a documented step, not a verbal habit

A disclosure that happens only in conversation is a disclosure you cannot prove later. The enforcement design under TRAIGA makes that proof matter. The Texas Attorney General can investigate and seek civil penalties for violations, and for a licensed physician the failure carries a second front: the licensing board. A missing disclosure carries two kinds of exposure at once: a fine risk and a professional-conduct risk.

That combination is why the durable answer is a recorded step, not a remembered courtesy. The same way informed consent and HIPAA notices became things a practice can show it provided, the AI-use disclosure needs to be something that lands in the chart with a date. When the question later becomes whether the patient was told before the point of care, the only safe answer is a record, not a recollection.

Building the disclosure into the encounter

Start with an honest inventory. Walk your clinical stack and mark every tool that participates in diagnosis or treatment, not the back-office software, but anything that shapes the clinical conclusion or the plan. Those are the encounters that now carry a disclosure obligation. Practices often discover the list is longer than expected once decision support and imaging tools are counted.

Then write one plain-language disclosure you can reuse and place it where the patient meets it at or before care. For many practices that means the intake packet, the pre-visit portal step, or the opening of the visit, with a fallback path for emergencies that documents the as-soon-as-reasonable disclosure after the fact. Add a charting field so a clinician or front-desk step records that it was given. Keep it short. A patient does not need a model card. They need to know AI was involved, told to them clearly, before the care happened.

The boundary, stated plainly

Because TRAIGA is one statute doing several jobs, it is worth marking the edge of this article. The employment-screening and governing-board obligations under TRAIGA are a separate matter covered elsewhere in this newsroom and are not the subject here. This piece is only the healthcare patient-disclosure rule: AI in diagnosis or treatment, disclosed to the patient, in plain language, at or before the point of care, effective January 1, 2026. For a Texas physician, that is the line that moved.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.