![\"1661154a.jpg\"\/](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2025\/08\/1661154a.jpg\")
<\/p>\nPurpose: The purpose of this tracker is to identify key
\nfederal and state health AI policy activity and summarize laws
\nrelevant to the use of AI in health care. The below reflects
\nactivity from January 1, 2025 through June 30, 2025. This is
\npublished on a quarterly basis.<\/p>\n
Activity on AI in health care has been a nationwide phenomenon
\nduring the 2025 legislative session: as of June 30, 2025, 46 states
\nhave introduced over 250 AI bills impacting health care and 17
\nstates passed 27 of those bills into law.<\/p>\n
See map below.1<\/p>\n
<\/p>\n
This year, so far, passed laws have primarily focused on three
\nkey areas:<\/p>\n
1. Use of AI-Enabled Chatbots: Actors across
\nthe health care ecosystem are integrating AI chatbots to improve
\nefficiency, enhance patient engagement, and expand access to care.
\nChatbot functionalities are being leveraged in administrative
\nfunctions (e.g., in support of patient scheduling) and in clinical
\nfunctions (e.g., early patient triage). States are taking action to
\nlegislate these tools in response to concerns that AI chatbots may
\nmisrepresent themselves as humans, produce harmful or inaccurate
\nresponses, or not reliably detect crises.<\/p>\n
Notably, five bills have been passed, and four signed into
\nlaw,2 related to AI-enabled chatbots this year. Of
\nthose, two directly address the use of chatbots in the delivery of
\nmental health services:<\/p>\n
Separately but related, Texas HB
Finally, Illinois HB
The other two laws that passed addressed concerns about
A dozen other chatbot bills were introduced but did not pass,
2. Payor Use of AI: Payors are increasingly
3. AI in Clinical Care: States are eager to
While most 2025 legislative sessions have now ended<\/a>, six states (CA, MA, MI, OH, PA, and WI)
In Q2, there was significant federal activity. While ultimately
In addition, federal agencies have solicited comment on AI and
For a summary of substantive federal action to date, see the
Health AI Laws Passed in 2025:<\/p>\n The below table represents the health AI laws that passed in
* Laws with an asterisk are those we consider “key state
State<\/p>\n
Summary<\/p>\n
Arizona*<\/p>\n
HB 2175<\/a> requires that a health care provider
Date Enacted: 5\/12\/2025<\/p>\n
Date Effective: 6\/30\/2026<\/p>\n
Maine*<\/p>\n
HP 1154<\/a> prohibits the use of artificial
Date Enacted: 6\/12\/2025<\/p>\n
Date Effective: 6\/18\/2025<\/p>\n
Maryland*<\/p>\n
HB 820<\/a> requires carriers (including health
Date Enacted: 5\/20\/2025<\/p>\n
Date Effective: 10\/1\/2025<\/p>\n
Montana<\/p>\n
HB 178<\/a> prohibits the AI use by government
Date Enacted: 5\/5\/2025<\/p>\n
Date Effective: 10\/1\/2025<\/p>\n
Nebraska*<\/p>\n
LB 77<\/a> establishes that AI algorithms may not be
Date Enacted: 6\/4\/2025<\/p>\n
Date Effective: 1\/1\/2026<\/p>\n
Nevada*<\/p>\n
AB 406<\/a> prohibits AI “providers” from
Date Enacted: 6\/5\/2025<\/p>\n
Date Effective: Upon passage and approval for the purpose of
New Mexico<\/p>\n
HB 178<\/a> establishes that the Board of Nursing
Date Enacted: 4\/8\/2025<\/p>\n
Date Effective: 6\/20\/2025<\/p>\n
New York*<\/p>\n
SB 3008<\/a> prohibits any person or entity to
Date Enacted: 5\/9\/2025<\/p>\n
Date Effective: 11\/5\/2025<\/p>\n
Oregon*<\/p>\n
HB 2748<\/a> mandates that “nonhuman”
Date Enacted: 6\/24\/2025<\/p>\n
Date Effective: 1\/1\/2026<\/p>\n
Texas*<\/p>\n
HB 149<\/a> sets requirements for government agency
Requirements for non-governmental developers and deployers of AI
An AI system deployed in relation to health care services or
Date Enacted: 6\/22\/2025<\/p>\n
Date Effective: 1\/1\/2026<\/p>\n
Texas*<\/p>\n
SB 815<\/a> prohibits a utilization review
Date Enacted: 6\/20\/2025<\/p>\n
Date Effective: 9\/1\/2025<\/p>\n
Texas*<\/p>\n
SB 1188<\/a> requires providers leveraging AI for
Date Enacted: 6\/20\/2025<\/p>\n
Date Effective: 9\/1\/2025<\/p>\n
Texas<\/p>\n
HB 3512<\/a> requires the Department of Information
Date Enacted: 6\/20\/2025<\/p>\n
Date Effective: 9\/1\/2025<\/p>\n
Utah*<\/p>\n
SB 226<\/a> repealed Utah SB 149 disclosure
Relatedly, in 2025, SB
Date Enacted: 3\/27\/2025<\/p>\n
Date Effective: 5\/7\/2025<\/p>\n
Utah*<\/p>\n
HB 452<\/a> requires suppliers of “mental
The law does not preclude chatbots from recommending that users
The Attorney General may impose penalties for violations of this
Finally, the law states that it is an affirmative defense to
Date Enacted: 3\/25\/2025<\/p>\n
Date Effective: 5\/7\/2025<\/p>\n
Other: State Activity Laws<\/p>\n
Over the past several decades, states have sought to understand
The following passed in 2025: Alabama HB
Key Federal Activity<\/p>\n
2025 Activity To-Date<\/p>\n
White House<\/p>\n
Congress<\/p>\n
Initial drafts of H.R. 1 included a ten-year moratorium on state
Bills to:<\/p>\n
Several others that touch on AI in health care, which we will
HHS Appointments and Announcements<\/p>\n
OCR<\/p>\n
ONC<\/p>\n
CMS<\/p>\n
FDA<\/p>\n
NIH<\/p>\n
DOJ<\/p>\n
Litigation continues over alleged use of AI to deny Medicare
FTC<\/p>\n
Footnotes<\/p>\n 1 This map does not include bills we categorize as
2 Illinois HB
3 Exceptions include: if the data is (a) requested by a
4 AI Companions defined as: “a system using
AI Companions exclude “any system used by a business
5 “Artificial Intelligence Systems” are defined
6 “Regulated occupations” are defined as
7 California SB 243<\/a>, New York AB
\nUtah’s HB 452<\/a>, effective as of 5\/7\/2025, requires
\n“mental health chatbots” to clearly and conspicuously
\ndisclose to the user that the chatbot is AI technology (and not a
\nhuman) at the beginning of any interaction, before the user
\naccesses features of the chatbot, and any time the user asks or
\notherwise prompts the chatbot about whether AI is being used. The
\nlaw prohibits the sale or sharing of individual user data from
\nmental health chatbots with any third party (with key clinical
\nexceptions3 ) or advertising a specific product or
\nservice during the conversation unless the chatbot clearly and
\nconspicuously identifies the advertisement and discloses any
\nsponsorships, business affiliations, or agreements that the
\nsupplier has with third parties to promote the product or service.
\nSeparately, Utah’s Office of Artificial Intelligence Policy has
\nrecently published best practices<\/a> for use of AI by mental health
\ntherapists.
\n
\nNew York’s enacted budget (SB
\n3008<\/a>), effective beginning 11\/5\/2025, is more broadly
\napplicable to “AI companions.”4 The law
\nprohibits any person or entity to operate or provide an “AI
\ncompanion” to someone in New York unless the model contains a
\nprotocol to take reasonable effort to detect and address suicidal
\nideation or expressions of self-harm expressed by the user. That
\nprotocol must, at a minimum: (1) detect user expressions of
\nsuicidal ideation or self-harm, and (2) refer users to crisis
\nservice providers (e.g., suicide prevention and behavioral health
\ncrisis hotlines) or other appropriate crisis services, when
\nsuicidal ideations or thoughts of self-harm are detected. AI
\ncompanion operators must provide a “clear and
\nconspicuous” notification\u2014either verbally or in
\nwriting\u2014that the user is not communicating with a human; that
\nnotification must occur at the beginning of any AI companion
\ninteraction, and at least every three hours after for continuous
\ninteractions.
<\/p>\n
\n149<\/a> (effective beginning 9\/1\/2026) is more broadly applicable
\nto “artificial intelligence systems”5 and
\nspecifically prohibits any development or deployment of these
\nsystems that would “incite or encourage” a person to
\ncommit physical self-harm, cause harm others, or engage in criminal
\nactivity. While the law is not specific to chatbots, these
\nprovisions may impact AI chatbots deployed in the state. The law
\nincludes several other broader provisions (see summary below).<\/p>\n
\n1806<\/a>\u2014which was sent to the Governor on June 24 (although
\nnot yet signed)\u2014prohibits the use of AI systems in therapy or
\npsychotherapy to make independent therapeutic decisions, directly
\ninteract with clients in any form of therapeutic communication, or
\ngenerate therapeutic recommendations or treatment plans without the
\nreview and approval by a licensed professional. This law, if
\nsigned, may substantially impair use of AI systems for mental
\nhealth services.<\/p>\n
\nmisrepresentation of chatbots as humans, as well as data use and
\nsharing: Maine HP 1154<\/a>, effective as of 6\/18\/2025,
\nprohibits chatbots that mislead or deceive users into believing
\nthey are interacting with a human, and Utah SB
\n226<\/a>, effective as of 5\/7\/2025, requires “regulated
\noccupations”6 to disclose use of AI in messaging in
\n“high-risk artificial intelligence interactions.”<\/p>\n
\nwhich were mainly general chatbot bills (not specific to health
\ncare) focused on setting forth disclosure requirements. There were
\nthree bills that did not pass that included provisions specific to
\nhealthcare chatbots and\/or had mental health specific
\nprovisions.7 We anticipate further activity in this area
\nnext legislative session.<\/p>\n
\nadopting AI across operations\u2014from utilization and quality
\nmanagement to fraud detection and claims adjudication\u2014with a
\nrecent NAIC survey finding that 84% of health insurers use AI or
\nmachine learning across their product lines.8 In
\nresponse, states are focusing on ways to mitigate potential harms
\nto beneficiaries from its use. Of the approximately 60 bills that
\nintroduced governing payer use of AI, only four passed. These four
\nlaws focus primarily on prohibiting the sole use of AI in making
\nmedical necessity denials or denying prior authorization and either
\nrequiring review by a physician of all AI decisions or prohibiting
\npayers from replacing physician\/peer review of medical
\nappropriateness with an AI tool. For more detail, see below Arizona
\n(HB 2175<\/a>), Maryland (HB
\n820<\/a>), Nebraska (LB
\n77<\/a>), and Texas (SB
\n815<\/a>).<\/p>\n
\nbalance AI’s potential to support clinical care with the
\npotential risks posed to patients and providers. This year, states
\nhave worked to establish guardrails on the use of AI in clinical
\ncare, such as provider oversight requirements, transparency
\nmandates, and safeguards against bias and misuse of sensitive
\nhealth data. Of the over 20 bills regulating provider use of AI
\nintroduced in 2025\u2014which included language regarding the role
\nof AI in clinical delivery, what provider oversight should be
\nrequired when using AI tools in clinical decision-making, and how
\nproviders should communicate AI use to patients\u2014four bills
\npassed. Texas HB 149<\/a> mandates that providers leveraging
\nAI systems for health care services or treatments provide
\ndisclosure to the patient (or their representative) no later than
\nthe date that the service or treatment is first provided (with
\nexception for emergency situations, when disclosure must be
\nprovided “as soon as reasonably possible”). Laws in
\nNevada (AB 406<\/a>) and Oregon (HB
\n2748<\/a>) prohibit AI systems from representing themselves as
\nlicensed providers; Nevada’s bill focused on an AI system
\nrepresenting itself as mental or behavioral health care providers
\nand Oregon’s on nurses. Texas SB
\n1188<\/a> requires providers leveraging AI for diagnostic or other
\npurposes to review information before entering it into patient
\nrecords.<\/p>\n
\nremain in session and are actively progressing legislation. We will
\ncontinue to track legislation in those states. See below
\ntable for a full summary of key health AI laws passed in 2025
\nand here<\/a> or a list of all AI laws
\npassed to-date.<\/p>\n
\nnot passed, a near final draft of H.R. 1 (“One Big Beautiful
\nBill”) included language that would have barred state or local
\ngovernments from enforcing laws or regulations on AI models or
\nsystems for up to ten years. More recently, the White House
\nreleased the “Winning the Race: America’s AI Action
\nPlan<\/a>,” which includes a significant deregulatory and
\ngeopolitical posture, and frames AI advancement as essential to
\nAmerican vitality, national security, and scientific leadership.
\nThe Plan, among other provisions, directs federal agencies to
\nidentify and repeal rules that could hinder AI development.
\nNotably, the plan declares that the “Federal government should
\nnot allow AI-related Federal funding to be directed towards states
\nwith burdensome AI regulations” and that Federal agencies with
\nAI-related discretionary funding consider a state’s AI
\nregulatory climate when making funding decisions. Separately, the
\nplan calls on the Federal Communications Commission (FCC) to
\nevaluate whether state AI laws interfere with its authorities to
\nregulate, among other things, communication and broadband-enabled
\nhealth technologies. To accelerate innovation, the plan proposes
\nthe establishment of regulatory sandboxes or AI Centers of
\nExcellence across the country\u2014supported by a variety of
\nfederal agencies (FDA, SEC, DOC)\u2014where developers can deploy
\nand test AI tools in real-world settings. In parallel, the
\nDepartment of Commerce (DOC), through NIST, would lead
\ndomain-specific efforts (e.g., health care, energy, and
\nagriculture) to develop national AI standards and evaluate
\nproductivity gains from AI in applied settings.<\/p>\n
\nhealth care: In May 2025, the Office of National Coordinator (ONC)
\nand Centers for Medicare & Medicaid (CMS) issued a request for information<\/a> seeking public
\nfeedback on digital tools, including AI, that can improve Medicare
\nbeneficiary access, improve interoperability, and reduce
\nadministrative burden. On June 27, CMS launched<\/a> a new model titled the Wasteful and
\nInappropriate Service Reduction (WISeR) Model to partner with
\ntechnology companies to “improv[e] and expedit[e]” prior
\nauthorization process compared to Original Medicare’s existing
\nprocesses and to reduce fraud for several services\/products that
\nCMS deemed at high risk of fraud or abuse in selected states. CMS
\nalso, in the CY 2026 Proposed Medicare Physician Fee Schedule,
\nrequested public comments on appropriate payment strategies for
\nsoftware as a service and AI.<\/p>\n
\ntable below.<\/p>\n
\n2025. For a full list of all laws prior to and including
\n2025, please see here<\/a>.<\/p>\n
\nlaws.” These are laws that, based on our review, are of
\ngreatest significance to the delivery and use of AI in health care
\nbecause they are broad in scope and directly touch on how health
\ncare is delivered or paid or because they impose significant
\nrequirements on those developing or deploying AI for health care
\nuse.<\/p>\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\n
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nindividually, exercising independent medical judgment, review
\nclaims and prior authorization requests prior to an insurer denying
\na claim or prior authorization. The law bans the sole use of any
\nother source to deny a claim or prior authorization.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nintelligence chatbots or similar technologies in trade and commerce
\nin a manner that may mislead or deceive consumers into believing
\nthey are interacting with a human being, unless the consumer is
\nclearly and conspicuously notified that they are not engaging with
\na human being.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\ninsurers, dental benefit plans, pharmacy benefit managers that
\nprovide utilization review, and any health benefit plans subject to
\nregulation by the state) to ensure that any AI tools used for
\nutilization review base decisions on medical\/clinical history,
\nindividual circumstances, and clinical information; does not solely
\nleverage group datasets to make decisions; does not “replace
\nthe role of a health care provider in the determination
\nprocess”; does not result in discrimination; is open for
\ninspection\/audit; does not directly or indirectly cause harm; and
\npatient data is not used beyond its intended use. The law mandates
\nthat AI tools may not “deny, delay or modify health care
\nservices.”<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nentities to “classify a person or group based on behavior,
\nsocioeconomic status, or personal characteristics resulting in
\nunlawful discrimination.” Requires government entities provide
\ndisclosures on any published material posted by AI not reviewed by
\na human.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nthe “sole basis” of a “utilization review
\nagent’s” (defined as any person or entity that performs
\nutilization review) decision to “deny, delay, or modify health
\ncare services” based whole or in part on medical necessity.
\nThe law requires utilization review agents to disclose use of AI in
\nutilization review process to each health care provider in its
\nnetwork, to each enrollee, and on its public website.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\n“explicitly or implicitly” indicating that an AI system
\nis capable of providing or is providing professional mental or
\nbehavioral health care. Prohibits providers of mental and
\nbehavioral health care from using or providing AI systems in
\nconnection to the direct provision of care to patients. Sets forth
\nthat providers may use AI tools to support administrative tasks
\nprovided that the provider must 1) ensure that use complies with
\nall applicable federal and state laws governing patient privacy and
\nsecurity of EHRs, health-related information, and other data,
\nincluding HIPAA, and 2) review the accuracy of any report, data, or
\ninformation compiled, summarized, analyzed, or generated by AI
\nsystems. The law requires the state agency to develop public
\neducation material focusing on, amongst other topics, best
\npractices for AI use by individuals seeking mental or behavioral
\nhealth care or experiencing a mental or behavioral health event.
\nAdditionally, the law prohibits all public schools (including
\ncharter schools or university schools) from using AI to
\n“perform the functions and duties of a school counselor,
\nschool psychologist, or school social worker” as related to
\nstudent mental health.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\nadopting any regulations and performing any other necessary
\npreparatory administrative tasks to carry out provisions of this
\nact. 7\/1\/2025 for all other purposes.<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nshall “promulgate rules establishing standards for the use of
\nartificial intelligence in nursing.”<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\noperate or provide an “AI companion” to someone in New
\nYork unless the model contains a protocol to take reasonable effort
\nto detect and address suicidal ideation or expressions of self-harm
\nexpressed by the user. Requires protocols to, at a minimum: (1)
\ndetect user expressions of suicidal ideation or self-harm, and (2)
\nrefer users to crisis service providers (e.g., suicide prevention
\nand behavioral health crisis hotlines) or other appropriate crisis
\nservices, when suicidal ideations or thoughts of self-harm are
\ndetected. Requires that AI companion operators provide a
\n“clear and conspicuous” notification\u2014either
\nverbally or in writing\u2014that the user is not communicating
\nwith a human; that notification must occur at the beginning of any
\nAI companion interaction, and at least every three hours after for
\ncontinuous interactions. Sets forth that the Attorney General has
\noversight authority and can impose penalties of $15,000\/day on an
\noperator that violates the law.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nentities, including AI tools, may not use the title of nurse or
\nsimilar titles, including advanced practice registered nurse,
\ncertified registered nurse anesthetist, clinical nurse specialist,
\nnurse practitioner, medication aide, certified medication aide,
\nnursing aide, nursing assistant, or certified nursing
\nassistant.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nand non-governmental use of AI. Requirements for government
\nagencies include: mandating that government agencies using AI
\nsystems that interact with consumers clearly and conspicuously
\ndisclose to each consumer, before or at the time of interaction,
\nthat the consumer is interacting with an AI system; prohibiting
\ngovernment entities from using AI systems that produce social
\nscoring, or developing or deploying an AI system that uses
\nbiometric identifiers to uniquely identify individuals if that use
\ninfringes on constitutional rights; and establishing an AI
\nRegulatory Sandbox Program and creates the “Texas Artificial
\nIntelligence Council.”<\/p>\n
\n
<\/p>\n
\ninclude: prohibiting deployers from deploying AI systems that aim
\nto “incite or encourage” a user to commit self-harm, harm
\nanother person, or engage in criminal activity and prohibiting
\ndevelopment or deployment of AI systems that discriminate.<\/p>\n
\n
<\/p>\n
\ntreatments must be disclosed by the provider to the recipient of
\nhealth services or their personal representative on the date of
\nservice, except in emergencies, when the provider shall disclose as
\nsoon as reasonably possible.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nagent’s use of an automated decision system (defined as an
\nalgorithm or AI that makes, recommends, or suggests certain
\ndeterminations) to “make, wholly or partly, an adverse
\ndetermination.” Adverse determinations are defined as
\ndeterminations that services are not medical necessary or
\nappropriate, or are experimental or investigational. Sets forth
\nthat the use of algorithms, AI, or automated decision systems for
\nadministrative support or fraud detection is allowable. Empowers
\nthe Commissioner of Insurance to audit and inspect use of
\ntools.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\ndiagnostic or other purposes to “review all information
\ncreated with artificial intelligence in a manner that is consistent
\nwith medical records standards developed by the Texas Medical
\nBoard.” In addition, a provider using AI for diagnostic
\npurposes must disclose the use of the technology to their
\npatients.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nResources to implement state-certified AI training programs
\n(similar to existing cybersecurity training protocols) for agency
\nstaff and local governments.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nprovisions and replaced them with disclosure requirements that are
\nsimilar, but required in more narrow scenarios. As with SB 149, the
\nlaw requires “regulated occupations” to prominently
\ndisclose that they are using computer-driven responses before they
\nbegin using generative AI for any oral or electronic messaging with
\nan end user. However, this disclosure is only required when the
\ngenerative AI is “high-risk,” which is defined as (a) the
\ncollection of personal information, including health, financial, or
\nbiometric data and (b) the provision of personalized
\nrecommendations that could be relied upon to make significant
\npersonal determinations, including medical, legal, financial, or
\nmental health advice or services.<\/p>\n
\n
<\/p>\n
\n332<\/a> passed, which extended the repeal date of SB 149 to July 1,
\n2027.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nhealth chatbots”9 to clearly and conspicuously
\ndisclose that the chatbot is AI technology and not a human at the
\nbeginning of any interaction, before the user access features of
\nthe chatbot and any time the user asks or otherwise prompts the
\nchatbot about whether AI is being used. Prohibits
\n“suppliers”10 of mental health chatbots
\nfrom:<\/p>\n
\n
\n
\nSelling or sharing individually identifiable health information
\nor user input with any third party, except if that information is
\n(a) requested by a health care provider with a user’s consent;
\n(b) provided to a health plan of a Utah user upon a user’s
\nrequest; or (c) shared by the supplier to ensure the effective
\nfunctionality of the tool, provided that the supplier and the
\nrecipient of such information comply with HIPAA regulations (as if
\nthe supplier were a covered entity and the other entity were a
\nbusiness associate).
\n
\n
\n
\nAdvertising a specific product or service during the
\nconversation unless the chatbot clearly and conspicuously
\nidentifies the advertisement as an advertisement and clearly and
\nconspicuously discloses any sponsorships, business affiliations, or
\nagreements that the supplier has with third parties to promote the
\nproduct or service. The law also prohibits any targeted
\nadvertisement based on the user’s input.
\n
\n
<\/p>\n
\nseek counseling, therapy, or other assistance, as necessary.<\/p>\n
\n
<\/p>\n
\nlaw.<\/p>\n
\n
<\/p>\n
\nliability under the law if the supplier demonstrates that they
\nmaintained documentation that describes development and
\nimplementation of the AI model that complies with the law and
\nmaintains a policy that meets a long list of requirements,
\nincluding ensuring that a licensed mental health therapist was
\ninvolved in the development and review process and has procedures
\nwhich prioritize user mental health and safety over engagement
\nmetrics or profit. In order for the affirmative defense to be
\navailable, the policy must be filed with the Division of Consumer
\nProtection.<\/p>\n
\n
<\/p>\n
\n
<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nAI technology before regulating it. For example, states have
\ncreated councils to study AI and\/or created AI-policy positions
\nwithin government in charge of establishing AI governance and
\npolicy. States have additionally tracked use of AI technology
\nwithin state agencies. These bills reflect states interest in the
\npotential role of AI across industries, and potentially in health
\ncare specifically.<\/p>\n
\n
<\/p>\n
\n365<\/a>, Kentucky SB 4<\/a>, Maryland SB
\n705<\/a>, Maryland HB 956<\/a>, Mississippi SB
\n2426<\/a>, Montana HJR 178<\/a>, New York SB
\n822<\/a>, Texas HB 149<\/a> (certain provisions), Texas SB
\n1964<\/a>, Texas HB 3512<\/a>, and West Virginia HB
\n3187<\/a>.<\/p>\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nTrump Administration released the “Winning the Race: America’s AI Action
\nPlan<\/a>,” which declared U.S. global dominance in AI a
\nnational imperative and outlines a comprehensive roadmap based on
\nthree key pillars: innovation, infrastructure, and international
\ndiplomacy.
\n
\n
\n
\nTrump Executive<\/a> Order<\/a> revoked Biden Administration’s AI
\nExecutive Order (Jan ’25).
\n
\n
\n
\nPolicies<\/a> on federal AI use and procurement
\n(April ’25).
\n
\n
\n
\nPublic comment<\/a> on AI Action Plan (concluded mid-March
\n’25).
\n
\n
\n
\nOMB issued a memo<\/a> focused on government adoption of AI
\nservices.
\n
\n
\n
\nGeneral de-regulatory approach and emphasis from administration
\non using AI to identify instances of fraud, waste, and abuse.
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nlegislation of AI. After much debate, this provision was struck
\nfrom final law on July 4, 2025.<\/p>\n
\n
<\/p>\n
\n
\n
\nEstablish<\/a> the National AI Research Resource
\ninitiative (Mar ’25).
\n
\n
\n
\nAllow<\/a> AI and machine learning to prescribe
\nmedication (Jan ’25).
\n
\n
\n
\nAllow<\/a> Medicare payment pathway for AI-enabled
\ndevices (April ’25).
\n
\n
<\/p>\n
\nreport on if they gain traction.<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nIn May 2025, U.S. Health & Human Services (HHS) designated
\nPeter Bowman-Davis acting Chief AI Officer at HHS.
\n
\n
\n
\nIn May 2025, Secretary Kennedy indicated<\/a> that HHS is already leveraging AI in
\nstandard operations, with attention to advancing novel
\ntreatments.
\n
\n
\n
\nThe week before inauguration in January, HHS announced
\nthey’d hired three executive positions: Chief AI Officer (Dr.
\nMeghan Dierks<\/a>), Chief Data Officer (Kristen Honey<\/a>), and Chief Technology Officer
\n(Alicia Rouault<\/a>), all central to Biden’s AI
\nstrategy roadmap. In mid-February, it was reported that all three
\nexecutives were on administrative leave. As of July 17, 2025, the
\nHHS Employee Directory<\/a> indicates Kristen Honey and
\nAlicia Rouault remain HHS employees (Chief Data Officer and Digital
\nServices Expert, respectively). Dr. Meghan Dierks left HHS in May
\n2025.
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nNon-Discrimination rule is subject to ongoing litigation; the
\nfirst Trump Administration reversed a prior version of the
\nrule.
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nIn May 2025, ONC and CMS issued a request for information<\/a> seeking public
\nfeedback on digital tools\u2014including AI\u2014that can improve
\nMedicare beneficiary access, improve interoperability, and reduce
\nadministrative burden.
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nCY2026 Proposed Medicare Physician Fee Schedule requested
\npublic comments on appropriate payment strategies for software as a
\nservice and artificial intelligence.
\n
\n
\n
\nOn June 27, CMS launched<\/a> a new model, the Wasteful and
\nInappropriate Service Reduction (WISeR) Model, to partner with
\ntechnology companies to use AI to “improv[e] and
\nexpedit[e]” prior authorization process compared to Original
\nMedicare’s existing processes to reduce fraud for several
\nservices\/products.
\n
\n
\n
\nIn final<\/a> rule for CY 2026, CMS chose not to
\nfinalize provisions regarding Medicare Advantage use of AI but
\nacknowledged the “broad interest” in AI and “will
\ncontinue to consider the extent to which it may be appropriate to
\nengage in future rulemaking in this area.”
\n
\n
\n
\nIn May 2025, ONC and CMS issued a request for information<\/a> seeking public
\nfeedback on digital tool\u2014including AI\u2014that can improve
\nMedicare beneficiary access, improve interoperability, and reduce
\nadministrative burden.
\n
\n
\n
\nUnder the Meaningful Measures 2.0<\/a> strategy, CMS is
\nprioritizing digital quality measures, including using AI to
\nidentify and address quality issues.
\n
\n
\n
\nDr. Mehmet Oz, the new administrator for CMS, has been reported
\nas promoting the use of artificial intelligence at CMS, in
\nparticular to combat fraud, waste and abuse<\/a>, and possibly
\nusing AI avatars instead of frontline health care workers as a way
\nto reduce costs without compromising quality.
\n
\n
\n
\nMACPAC presented study<\/a> findings on the use of AI in prior
\nauthorization processes in Medicaid (Feb ’25)
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nDraft guidance<\/a> for developers of AI-enabled medical
\ndevices.
\n
\n
\n
\nIn June 2025, launched “Elsa<\/a>” to support departmental
\nefficiency
\n
\n
\n
\nIn May 2025, Jeremy Walsh was hired<\/a> as FDA’s chief AI officer and head
\nof IT.
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
<\/p>\n
\nAdvantage claims. In June 2025, DOJ announced charges<\/a> against over 300 defendants
\nfor participation in health care fraud schemes, with a parallel
\nannouncement from CMS on the successful prevention of $4 billion in
\npayments for false and fraudulent claims.<\/p>\n
\n
\n
\n
\n
\n
<\/p>\n
\n
\n
\n
\nNo significant activity. An FTC publication<\/a> noted the department’s focus
\non potential AI harms and also reinforced that existing laws apply
\nto AI technologies.
\n
\n
\n
\n
<\/p>\n
\n“Other: State Activity Laws,” which generally are bills
\nthat create councils or tasks forces to study AI or are related to
\nnarrow state activities.<\/p>\n
\n1806<\/a> passed and was sent to the Governor on June 24 , but, as
\nof the time of publication has not yet been signed.<\/p>\n
\nhealth care provider with a user’s consent; (b) provided to a
\nhealth plan of a Utah user upon a user’s request; or (c) shared
\nby the supplier to ensure the effective functionality of the tool,
\nprovided that the supplier and the recipient of such information
\ncomply with HIPAA (even if not a covered entity or business
\nassociate).<\/p>\n
\nartificial intelligence, generative artificial intelligence, and\/or
\nemotional recognition algorithms designed to simulate a sustained
\nhuman or human-like relationship with a user by:<\/p>\n
\nretaining information on prior interactions or user
\nsessions and user preferences to personalize the interaction and
\nfacilitate ongoing engagement with the AI companion;
\n
\nasking unprompted or unsolicited emotion-based questions
\nthat go beyond a direct response to a user prompt; and
\n
\nsustaining an ongoing dialogue concerning matters
\npersonal to the user.”
<\/p>\n
\nentity solely for customer service or to strictly provide users
\nwith information about available commercial services or products
\nprovided by such entity, customer service account information, or
\nother information strictly related to its customer service; any
\nsystem that is primarily designed and marketed for providing
\nefficiency improvements or, research or technical assistance; or
\nany system used by a business entity solely for internal purposes
\nor employee productivity.”<\/p>\n
\nas “any machine-based system that, for any explicit or
\nimplicit objective, infers from the inputs the system receives how
\nto generate outputs, including content, decisions, predictions, or
\nrecommendations, that can influence physical or virtual
\nenvironments.”<\/p>\n
\noccupations regulated by the Department of Commerce that require an
\nindividual to obtain a license or state certification to practice
\nthe occupation. “High-risk artificial intelligence
\ninteractions” are defined as interactions with generative
\nartificial intelligence that involve 1) the collection of sensitive
\npersonal information (including health data, financial data, or
\nbiometric data); 2) the provision of personalized recommendations,
\nadvice, or information that could be reasonably relied upon to make
\nsignificant personal decisions (including medical advice or
\nservices or mental health advice or services, among others), or 3)
\nother applications as defined by division rule.<\/p>\n
\n6767<\/a>, and North Carolina SB
\n624<\/a> were all introduced in 2025 and contain provisions focused
\non AI chatbots and mental health.<\/p>\n