Texas Made AI Disclosure a Duty in January. The Complaint Form Opened September 1.
Business and Commerce Code Sec. 552.051(f) has required every provider of health care service or treatment in Texas to tell patients when an artificial intelligence system is used in relation to that treatment since January 1, 2026. As of September 1 the attorney general's office is taking complaints about it on a form of its own.
On September 1, 2026, the Texas attorney general's office added an artificial intelligence complaint form to its public consumer complaint page, alongside the general and privacy forms. The office describes it as being for “concerns that an artificial intelligence (AI) system is being developed or deployed in a way that violates Texas law.”
The duty it collects complaints against has been in force since January 1.
The subsection that reaches a private practice
The Texas Responsible Artificial Intelligence Governance Act reads, for most of its length, like a law written for technology companies and state agencies. Section 552.051(b) is addressed to governmental agencies. Then subsection (f) widens it:
“If an artificial intelligence system is used in relation to health care service or treatment, the provider of the service or treatment shall provide the disclosure under Subsection (b) to the recipient of the service or treatment or the recipient's personal representative not later than the date the service or treatment is first provided, except in the case of emergency, in which case the provider shall provide the required disclosure as soon as reasonably possible.”
Subsection (a) defines the services in question as those “provided by an individual licensed, registered, or certified under applicable state or federal law to provide those services.” A behavior analyst licensed by the Texas Department of Licensing and Regulation under Occupations Code Chapter 506 is that individual. Nothing in subsection (f) limits it to public providers.
What the disclosure has to say, and when
Subsection (f) borrows its content from subsection (b): the disclosure tells the recipient they are interacting with an artificial intelligence system. Three more clauses set the terms.
- It is due no later than the first date of service — not the first calendar year, not at renewal. Emergencies get “as soon as reasonably possible.”
- Obviousness is not a defense. Subsection (c): “A person is required to make the disclosure under Subsection (b) regardless of whether it would be obvious to a reasonable consumer that the consumer is interacting with an artificial intelligence system.”
- It must be clear and conspicuous, in plain language, and free of dark patterns as that term is defined in Section 541.001. Subsection (e) permits a hyperlink to a separate web page.
The definition of the thing being disclosed is the widest part of the act. Section 551.001(1): “any machine-based system that, for any explicit or implicit objective, infers from the inputs the system receives how to generate outputs, including content, decisions, predictions, or recommendations, that can influence physical or virtual environments.” That is drafted to cover a note-drafting assistant and a scheduling engine that recommends, not only a chatbot a parent talks to.
What enforcement looks like
The attorney general has exclusive authority under Section 552.101, and the same section states the chapter “does not provide a basis for, and is not subject to, a private right of action.” No family sues you over this. The state does, or nobody does.
- Sixty days to cure from the written notice of violation, under Section 552.104, with documentation of the cure submitted to the attorney general.
- Curable violations: $10,000 to $12,000 each. Uncurable: $80,000 to $200,000 each. Continued violations: $2,000 to $40,000 a day.
- An affirmative defense exists. Section 552.105(e) covers a defendant who substantially complies with the current NIST “Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile” or another nationally or internationally recognized risk management framework.
The limits
The trigger and the message do not quite line up, and that gap is unresolved. Subsection (f) fires when an AI system is used “in relation to” health care service or treatment. The disclosure it then requires is subsection (b)'s — that the recipient is interacting with an AI system. A tool that drafts your session notes is used in relation to treatment and is not something the family interacts with at all. The statute does not say which reading governs, and the attorney general has not published guidance on it.
The licensing side is still open too. TDLR proposed changes to the behavior analyst rules at 16 TAC Chapter 121 in the Texas Register on June 19, 2026, one of which would set “ethical requirements for behavior analysis licensees who use AI systems in client services.” Comments closed July 20, 2026. As of this writing the department's behavior analyst news page shows no adoption, so the license-side rule is proposed and not in effect. The statutory duty does not wait for it.
A separate 2025 Texas act, SB 1188, effective September 1, 2025, addresses practitioner review of AI-generated records against standards set by the Texas Medical Board. Whether its practitioner definition reaches licensed behavior analysts is not settled on the face of the summaries we read, and we have not confirmed it either way.
What you must know or do
- List every system in your clinical path that turns inputs into content, decisions, predictions or recommendations. Note drafting, session-data summarizing, intake triage, authorization letter drafting, scheduling that suggests rather than stores. Section 551.001(1) is written broadly enough that “it only summarizes” is not an answer to it.
- If any of them touches a Texas patient's care, the disclosure is due before or on the first date of service. That means the intake packet, not the annual notice of privacy practices. Families already in treatment on January 1 are the population to sort out first.
- Write it plainly and put it where it is found. Subsection (e) allows a hyperlink, so a one-page notice linked from the intake form meets the form requirement. A line buried in a consent bundle does not meet “clear and conspicuous.”
- Do it even where it is obvious. Subsection (c) exists specifically to remove that argument, and it is the cheapest violation to avoid.
- Owners: decide now whether you can claim 552.105(e). Using a NIST-aligned vendor is not the same as documenting your own substantial compliance with the framework. If you intend to rely on the defense, the documentation has to exist before the notice arrives, and the notice starts a 60-day clock.