AI in the Front Office

State Laws on AI in Patient Communications

California, Texas and Utah rules on disclosing AI in patient calls, messages and chat, the federal layer on top, and a multi-state compliance approach.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing7 min read

On this page
  1. Why states are writing these rules now
  2. Three laws at a glance
  3. California: disclaimers on clinical messages
  4. Texas: disclosure when AI is part of care
  5. Utah: high-risk interactions and direct questions
  6. The federal layer on top
  7. A multi-state approach that holds up
  8. Common misreadings to avoid
  9. Example: a fictional telehealth group
  10. Frequently asked questions
  11. References
AI in the Front Office: State Laws on AI in Patient CommunicationsArticle coverAI IN THE FRONT OFFICEState Laws on AI inPatient CommunicationsAva Medical IntakeInsightsAI in the Front Office: State Laws on AI in Patient CommunicationsArticle coverAI IN THE FRONT OFFICEState Laws on AI inPatientCommunicationsAva Medical Intake

Key takeaways

  • California requires an AI disclaimer and human-contact instructions on generative AI communications about clinical information, unless a licensed provider reviews them[1].
  • Texas requires health care providers to disclose AI used in relation to a service or treatment, by the date it is first provided, from January 1, 2026[2].
  • Utah requires regulated occupations to disclose generative AI prominently in high-risk interactions, including those collecting health data or giving medical advice[3].
  • With few practices holding a formal AI policy[4], the simplest compliant habit is to disclose AI everywhere, keep it out of clinical judgment and log human review.

Why states are writing these rules now

AI has moved from pilot to routine in outpatient care. In MGMA polling, AI played some role in patient visits at 83% of US medical groups in 2026, up from 71% a year earlier[5]. Much of that use touches patients directly: voice assistants on phone lines, AI-drafted portal replies, chat on websites and automated texts.

US medical groups where AI plays some role in patient visitsMGMA Stat poll, August 2026, 189 responses
US medical groups where AI plays some role in patient visitsColumn chart. 2025: 71%; 2026: 83%.0%25%50%75%100%2025: 71%71%20252026: 83%83%2026US medical groups where AI plays some role in patient visitsColumn chart. 2025: 71%; 2026: 83%.0%25%50%75%100%2025: 71%71%20252026: 83%83%2026

Source: Medical Group Management Association (MGMA), 2026[5]

Show the data
ItemValue
202571%
202683%

State legislatures have responded with transparency rules rather than bans. The common thread is that patients should know when they are dealing with AI and how to reach a human. Practice readiness lags: only 20% of medical groups report having AI governance or a formal AI-use policy, and 56% have none[4]. This article summarizes three state laws that already affect patient communications. It is not legal advice, rules change, and other states have bills in progress, so confirm your position with counsel.

Three laws at a glance

Summary of statute text as published by each legislature. Check the current text before relying on it.
ItemCaliforniaTexasUtah
LawHealth and Safety Code § 1339.75 (AB 3030)H.B. 149, Texas Responsible Artificial Intelligence Governance ActS.B. 226 (2025), Utah Code 13-75
WhoHealth facilities, clinics, physician offices, group practicesThe provider of the health care service or treatmentIndividuals providing services in a regulated occupation
TriggerGenerative AI creates written or verbal communications about patient clinical informationAn AI system is used in relation to a health care service or treatmentGenerative AI used in a high-risk interaction, such as collecting health data or giving medical advice
RequirementDisclaimer that AI generated it, plus instructions to reach a human providerDisclosure to the patient or representative, clear, conspicuous, plain language, no dark patternsProminent disclosure: verbally at the start of a spoken interaction, in writing before a written one
Key exceptionCommunications read and reviewed by a licensed or certified provider; administrative matters such as scheduling and billingEmergencies: disclose as soon as reasonably possibleApplies to high-risk interactions; separate duty to answer if a consumer clearly asks
In forceBill from the 2023–24 session; in forceJanuary 1, 2026May 7, 2025

California: disclaimers on clinical messages

California's rule targets generative AI used to produce patient communications that relate to a patient's health status. The statute defines patient clinical information that way and expressly excludes administrative matters, including appointment scheduling, billing and other clerical or business matters[1]. A voice assistant that only books visits is therefore outside the core requirement, while an AI-drafted message explaining lab results is inside it.

Where the rule applies, placement is specific. Written communications such as letters and emails carry the disclaimer prominently at the beginning. Continuous online interactions, including chat-based telehealth, display it throughout. Audio communications state it verbally at the start and the end, and video displays it throughout. Each communication must also give clear instructions on how to contact a human health care provider. If a licensed or certified provider reads and reviews the AI-generated communication, the requirements do not apply[1].

  • Map which of your AI outputs touch clinical information: results explanations, medication messages, symptom replies.
  • For each, decide whether a licensed provider will review before sending, and record the review.
  • Where no review happens, add the disclaimer in the required position and a human-contact line.
  • Do not assume scheduling and billing tools are covered or exempt in every case; ask counsel about messages that mix both.

Texas: disclosure when AI is part of care

The Texas Responsible Artificial Intelligence Governance Act took effect on January 1, 2026. Its disclosure section requires government agencies to tell consumers when they are interacting with an AI system, and extends a disclosure duty to health care: if an AI system is used in relation to a health care service or treatment, the provider must disclose it to the recipient or their personal representative no later than the date the service or treatment is first provided. In an emergency, the disclosure is due as soon as reasonably possible[2].

The manner rules are short but strict. The disclosure must be clear and conspicuous, written in plain language, and must not use a dark pattern[2]. In practice, that points toward a plain notice delivered at registration or the first contact, repeated in channels where AI speaks directly to patients, rather than a line buried in a consent packet.

Utah: high-risk interactions and direct questions

Utah's 2025 amendments, codified in Utah Code 13-75, contain two duties. First, a supplier using generative AI in a consumer transaction must disclose that the person is dealing with AI, not a human, if the person clearly and unambiguously asks. Second, an individual providing services in a regulated occupation must prominently disclose generative AI use when it is a high-risk interaction[3].

High-risk interactions include those that collect sensitive personal information, such as health data, and those that provide personalized recommendations or information that could be relied on for significant personal decisions, including medical advice or services. The disclosure must be given verbally at the start of a spoken interaction and in writing before a written one[3]. A front office voice that asks for symptoms or medication details would likely sit closer to the high-risk line than one that confirms an appointment time.

The federal layer on top

State rules sit alongside federal ones. Section 1557 rules require covered health entities to make reasonable efforts to identify and mitigate discrimination risks from patient care decision support tools, including AI tools that use factors such as race, national origin, sex, age or disability[6]. For outbound calls, the FCC ruled on February 8, 2024 that calls made with AI-generated voices count as artificial under the Telephone Consumer Protection Act[7], which brings consent and identification rules with them. See the FCC ruling on AI voices for details.

A multi-state approach that holds up

  1. List where your patients are

    Telehealth and multi-site groups often serve patients in several states. The patient's location can matter as much as the practice's.

  2. Inventory AI touchpoints

    Phone greetings, overflow lines, reminders, text replies, website chat, portal drafts and any outbound AI calls.

  3. Classify each one

    Administrative only, clinical information, or collecting health data. The classification drives which rules apply.

  4. Adopt the strictest common standard

    Disclose AI at the start of every interaction, give a human route, and keep AI out of clinical judgment. That satisfies the core of all three laws in most cases.

  5. Log human review

    Where a licensed provider reviews AI drafts, record who and when. The record is your evidence under California's exemption.

  6. Assign a watcher

    Name one person to track new state laws each quarter and update the policy.

Common misreadings to avoid

State AI rules are short, which makes them easy to misread. These are the mistakes practice teams most often make when they first read them, and the safer interpretation in each case.

  • Assuming the law only applies to chatbots. California covers verbal communications too, and Texas applies to any AI system used in relation to a health care service, not just conversational tools.
  • Assuming a clinician's name on the message counts as review. California's exemption depends on a licensed or certified provider actually reading and reviewing the AI-generated communication, so build a step that proves it happened.
  • Assuming the practice's home state is the only one that matters. Telehealth, patients who travel and multi-state groups can bring other states' rules into play.
  • Treating a privacy policy line as disclosure. Texas asks for a clear and conspicuous disclosure in plain language, and Utah asks for a prominent one at the start of the interaction.
  • Thinking disclosure settles everything. These laws sit on top of existing duties: licensing board standards, HIPAA, call recording consent and federal anti-discrimination rules all still apply.
  • Waiting for one national standard before acting. The state rules described here are already in force, and a federal rule, if one comes, may not replace them.

When in doubt, ask two questions about any AI touchpoint. Would a reasonable patient know, before sharing anything, that software is involved? And can that patient reach a person without friction? If the answer to both is yes, you are aligned with the purpose of every law discussed here, even as details differ.

Put the answers in writing. A one-page register of your AI touchpoints, their classification, the disclosure wording used and the date each was last checked gives counsel what they need, and gives staff a single reference when a patient asks.

Example: a fictional telehealth group

Example: a fictional behavioral health group based in Utah sees patients by video in Utah, Texas and California. It uses an AI assistant to schedule, and AI drafts of follow-up messages that clinicians edit. Counsel advises that scheduling calls are administrative, but intake calls that ask about symptoms involve health data. The group changes its intake design so the assistant schedules only, with symptom screening done by a clinician on video.

Example, continued: it adds a single disclosure to every AI touchpoint, verbal at the start of calls and written at the top of chats, and requires clinicians to review every AI-drafted clinical message, recording the reviewer. The result is one process that works across all three states rather than three separate ones.

Frequently asked questions

Which states require AI disclosure in health care?

California, Texas and Utah have laws affecting patient communications: California for AI-generated clinical messages, Texas for AI used in health care services from January 1, 2026, and Utah for high-risk interactions in regulated occupations. Other states are considering bills, so check with counsel.

Does California's AB 3030 apply to AI appointment scheduling?

The statute excludes administrative matters such as appointment scheduling and billing from patient clinical information, so pure scheduling is generally outside its core requirement. Messages that mix scheduling with clinical content need closer review.

What is the simplest way to comply across several states?

Disclose AI at the start of every interaction, give patients a route to a human, keep AI out of clinical decisions, and log licensed provider review of any AI-drafted clinical message.

References

  1. California Legislature (code text via California Public Law) (2024). California Health and Safety Code § 1339.75 (added by AB 3030) (opens in a new tab) ↑

  2. Texas Legislature (2025). H.B. 149, Texas Responsible Artificial Intelligence Governance Act (enrolled text) (opens in a new tab) ↑

  3. Utah State Legislature (2025). S.B. 226, Artificial Intelligence Consumer Protection Amendments (enrolled) (opens in a new tab) ↑

  4. Medical Group Management Association (MGMA) (2026). AI governance in medical group practices: Rules for the humans in the loop (opens in a new tab) ↑

  5. Medical Group Management Association (MGMA) (2026). As AI use expands, medical practice leaders shift attention to measurement (opens in a new tab) ↑

  6. Legal Information Institute, Cornell Law School (Electronic Code of Federal Regulations text) (current). 45 CFR § 92.210 - Nondiscrimination in the use of patient care decision support tools (opens in a new tab) ↑

  7. Federal Communications Commission (2024). FCC Makes AI-Generated Voices in Robocalls Illegal (news release) (opens in a new tab) ↑

Book a demo

See Ava handle your calls.

Book a short demo. We’ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo