State laws on AI therapy, mental health chatbots and companion chatbots, in plain words. Most of these laws put duties on the companies that build or run chatbots. The therapy bans go further and also limit how licensed professionals may use AI. Each card says whether the law reaches addiction counselors and whether it exempts peer support.
Last verified
Four kinds of law
Therapy ban
AI may not provide therapy, and licensed professionals face limits on how they use AI.
Professional impersonation
AI may not be advertised or presented as a licensed health professional, including a mental health professional.
Companion chatbot safety
Rules for AI companions and other chatbots that hold open-ended conversations with the public. Most require a plan for responding to suicide and self-harm, and many add protections for minors.
Disclosure
Chatbots must tell people they are talking with AI, not a human.
We list only laws we confirmed on official state websites, and each one links to its official text. A state that is not listed may still have a law we have not found. This page is not legal advice. If you need to know how a law applies to your own practice or program, ask a lawyer licensed in that state.
California
Companion chatbot safety
SB 243 (Statutes of 2025, ch. 677) Companion chatbots
A companion chatbot must say it is not human when a person could be misled. It must keep and publish a protocol that refers users who express suicidal thoughts or self-harm to crisis services.
Signed
Effective
. Annual reports to the state Office of Suicide Prevention begin July 1, 2027.
Covers addiction counselors?
No. It applies to whoever runs a companion chatbot, whatever their profession, and does not regulate counselors' own practice. The text does not mention substance use or addiction.
Peer-support exemption
Not addressed.
Enforced by
People who are harmed can sue (a private right of action). No state agency is named.
Penalties
The greater of actual damages or $1,000 per violation, plus attorney's fees and costs.
Note
SB 1119 (below) deletes this law's separate rules for users the operator knows are minors, starting January 1, 2027 (the date SB 1119 takes effect; the law does not print it). SB 1119's own child-safety duties apply from July 1, 2027.
AB 489 (Statutes of 2025, ch. 615) Health care professions: deceptive terms or letters: artificial intelligence
An AI system's advertising or features may not use words, letters or phrases suggesting that its care, advice or assessments come from a licensed health professional.
Signed
Effective
Covers addiction counselors?
Unclear. It covers the health care professions licensed under Division 2 of California's Business and Professions Code. It has no wording specific to substance use.
Peer-support exemption
Not addressed.
Enforced by
The relevant California health care licensing board or enforcement agency, which may ask a court to stop the violation.
Penalties
Each use counts as a separate violation. The title-protection laws it extends are crimes, so the Legislature says it expands existing crimes.
A companion chatbot that allows children must be checked for risks to children before release. If the company determines there is a credible and imminent threat that a child will engage in suicide or self-harm, it must notify a parent whose account is linked to the child's (unless that would put the child at risk of serious harm) or give the child a direct way to reach 988 or a similar crisis line. Default settings must protect children, including time limits, and the company must take reasonable steps so the chatbot does not encourage a child toward self-harm or “consumption of narcotics or alcohol.”
Signed
Effective
. The law does not print this date. It follows from California's rule that regular statutes take effect on January 1 of the next year. The main child-safety duties apply from July 1, 2027.
Covers addiction counselors?
No. It applies to whoever runs a companion chatbot that children can use, and does not regulate counselors' own practice. It requires reasonable steps to keep a chatbot from encouraging a child toward “consumption of narcotics or alcohol,” and from trying to diagnose or treat a child unless the chatbot is designed for that purpose and is regulated by the FDA as a medical device and under HIPAA.
Peer-support exemption
Not addressed.
Enforced by
Public prosecutors can sue for penalties; only the Attorney General may seek an injunction. A child harmed by certain violations, or the child's parent or guardian, may also sue.
Penalties
Up to $5,000 per affected child for a negligent violation, and up to $15,000 per affected child for an intentional one.
HB 26-1195 (Session Laws of Colorado 2026, ch. 358) An Act concerning restrictions on the use of artificial intelligence relating to psychotherapy services
No one may advertise or present an AI system as providing psychotherapy, as equivalent to services from a Colorado mental health licensee, or as keeping a user's data confidential the way a therapist does. Licensed professionals may use AI for administrative or supplementary support if they review its outputs. AI may take part in therapeutic communication with clients only while the professional takes part in real time, and may not generate treatment recommendations or plans without the professional's review and approval. Clients must get written notice and give written consent before AI records or transcribes a session.
Signed
Effective
Covers addiction counselors?
Yes. Its limits apply to everyone licensed, registered or certified to provide psychotherapy under article 245 of title 12, which covers licensed and certified addiction counselors.
Peer-support exemption
Not addressed. The law does not mention peer support. It exempts wellness tools, such as self-help, coaching, journaling and safety planning, that do not diagnose or treat mental health disorders and clearly say they are not a substitute for clinical care.
Enforced by
The licensing board, for professionals' use of AI. The advertising and claim rules are part of the Colorado Consumer Protection Act.
Note
The advertising and claim rules sit in part 17 of the Colorado Consumer Protection Act. SB 26-189 (2026) repeals and re-enacts part 17, effective January 1, 2027, and 2026 laws that add to part 17 collide on section numbers (two different sections are numbered 6-1-1708). How they fit together will show only when the 2026 Colorado Revised Statutes are published, so we do not list penalties yet.
HB 26-1263 (Session Laws of Colorado 2026, ch. 208) An Act concerning requirements for an operator of a conversational artificial intelligence service
A conversational AI service must tell users it is AI at the first interaction each day, at least every three hours or with a standing notice, and whenever asked if it is human. It must keep a suicide and self-harm protocol that refers users to crisis services, not law enforcement. It may not claim its output comes from, or equals the services of, a licensed health care, legal or mental health professional or a dietitian. Users it knows are minors get added protections.
Signed
Effective
. Its duties apply from January 1, 2027. Annual reports to the Attorney General begin July 1, 2027.
Covers addiction counselors?
Unclear. It applies to operators of conversational AI services and does not regulate counselors' own practice. It bars claims that output comes from a “licensed, certified, or registered mental health professional” but does not define that term, so it does not say whether certified addiction counselors are included.
Peer-support exemption
Not addressed. The law does not mention peer support. It excludes services used by or for a HIPAA-covered entity or its business associate.
Note
This law's rules sit in part 17 of the Colorado Consumer Protection Act. SB 26-189 (2026) repeals and re-enacts part 17, effective January 1, 2027, and SB 26-189 and this law each add a different section numbered 6-1-1708, so one will be renumbered. How they fit together will show only when the 2026 Colorado Revised Statutes are published, so we cite this law by its session-law chapter and do not list the enforcer or penalties yet.
SB 5 (Public Act 26-15, sections 4 to 6) An Act Concerning Online Safety
An AI companion must keep and post a protocol that detects expressions of suicide, self-harm or imminent violence and refers users to resources including 988, and the operator must take reasonable measures so it never claims to be human. It must tell users it is not human when a reasonable person could think otherwise. For users known or suspected to be minors, it must take measures against encouraging self-harm or “the unlawful consumption of alcohol or drugs,” and against offering mental health services unless it is designed for that and meets added conditions.
Signed
Effective
. This is the date for sections 4 to 6, the AI companion sections. Other parts of the act have other dates.
Covers addiction counselors?
No. It applies to whoever runs an AI companion, whatever their profession, and does not regulate counselors' own practice. Its term “licensed mental health professional” includes licensed alcohol and drug counselors.
Peer-support exemption
No. Its definition of mental health service includes “peer or recovery supports.” It excludes AI used solely for treatment-support functions, such as medication reminders, that does not present as human and is not designed to meet a user's social or emotional needs.
Enforced by
Connecticut Attorney General only. No private right of action.
Penalties
Up to $5,000 per wilful violation, and up to $25,000 per violation of a court order.
SB 540 (Act 518) Online Internet Safety; certain disclosures related to conversational AI services; require
An AI companion chatbot must say it is AI, not a person, at the start of each interaction and at least every three hours (every hour for minors). It must have a protocol that refers users to the 988 Suicide and Crisis Lifeline or comparable crisis services. It may not be programmed to claim it is licensed or authorized to provide mental health, behavioral health, medical or counseling services unless the operator is. Minors get added protections.
Signed
Effective
Covers addiction counselors?
No. It applies to operators of AI companion chatbots and does not regulate counselors' own practice. Its ban on false claims covers behavioral health and counseling services in general and does not mention substance use.
Peer-support exemption
Not addressed.
Enforced by
Georgia Attorney General.
Penalties
Up to $10,000 per knowing violation, plus compensatory damages, costs and attorney's fees. Each day counts as a separate violation for each affected user.
Note
The text is the As Passed version the Governor's Office publishes as signed; no signature page is posted.
An AI companion must say it is AI when a person could think it is human, and must keep a suicide and self-harm protocol that refers users to crisis services. It may not claim to be designed to provide professional mental or behavioral health care.
Signed
Effective
. Annual reports to the state Department of Health begin January 1, 2028.
Covers addiction counselors?
Unclear. It does not mention substance use. It bars claims of providing “professional mental or behavioral health care” but does not say whether that includes addiction counseling.
Peer-support exemption
Not addressed.
Enforced by
The Hawaii Attorney General or the Office of Consumer Protection. No private right of action.
Penalties
$500 to $10,000 per violation; each day counts as a separate violation.
S 1297 (2026 Session Laws, ch. 249) Conversational AI Safety Act
A conversational AI service must say it is AI when a person could think it is human, and must keep a protocol that refers users who express suicidal thoughts to crisis services. It may not be programmed to claim it is designed to provide professional mental or behavioral health care. Minor account holders get added protections.
Signed
Effective
Covers addiction counselors?
No. It applies to operators of conversational AI services and does not regulate counselors' own practice. Its claim ban covers “professional mental or behavioral health care” in general and does not mention substance use.
Peer-support exemption
Not addressed.
Enforced by
Idaho Attorney General. No private right of action.
Penalties
The greater of actual damages or $1,000 per violation, capped at $500,000 per operator, plus an injunction.
HB 1806 (Public Act 104-0054) Wellness and Oversight for Psychological Resources Act
No one may provide or advertise therapy in Illinois, including through AI, unless a licensed professional provides it. Licensed professionals may use AI only for administrative or supplementary support, never to make independent therapy decisions, talk with clients as a form of therapy, or detect emotions or mental states.
Signed
Effective
Covers addiction counselors?
Yes. Its list of licensed professionals includes a certified alcohol and other drug counselor authorized to provide therapy.
Peer-support exemption
Yes. Peer support is exempt. The law defines it as services from people with lived experience of mental health conditions or “recovery from substance use” that offer encouragement, understanding and guidance “without clinical intervention.” The definition describes services from people; the law does not say whether an AI tool can count as peer support.
Enforced by
Illinois Department of Financial and Professional Regulation (IDFPR). The law does not provide for lawsuits by individuals.
SF 2417 (2026 Iowa Acts, ch. 1068) Conversational Artificial Intelligence Services — Limitations, Required Disclosures, and Civil Penalties
Conversational AI services must tell minors, and anyone who could think they are human, that they are AI. They must keep a suicide and self-harm protocol and add protections for minors. They may not be programmed to suggest they provide psychology or behavioral health services that require an Iowa license.
Signed
Effective
. The law is on the books from July 1, 2026, but its rules apply starting July 1, 2027.
Covers addiction counselors?
Unclear. It does not mention substance use. It refers to services that need a license under Iowa Code chapters 154B (psychology) and 154D (mental health professions).
Peer-support exemption
Not addressed.
Enforced by
Iowa Attorney General. No private right of action.
Penalties
The greater of actual damages or $1,000 per violation, capped at $500,000 per operator.
LD 2082 (Public Law 2025, ch. 687) An Act to Regulate the Use of Artificial Intelligence in Providing Certain Mental Health Services
No one may provide, advertise or offer therapy or psychotherapy to the public, including through AI, unless a licensed professional provides it. Licensed professionals may use AI only for administrative or supplementary support, and supplementary support needs a recorded or transcribed session with written notice and consent. AI may not make independent therapeutic decisions, interact directly with clients in therapeutic communication, or generate treatment plans without the professional's review and approval.
Signed
Effective
Covers addiction counselors?
Yes. Its definition of licensed professional includes alcohol and drug counselors licensed or certified under Title 32, chapter 81 and authorized to provide therapy, and it adds the same AI limits to that chapter.
Peer-support exemption
Not addressed. The law does not mention peer support. Its only exception is AI used solely in a research project approved by an institutional review board.
Enforced by
The Maine Attorney General, under the Unfair Trade Practices Act, and consumers who lose money or property can sue. Licensing boards discipline licensed professionals.
Penalties
Up to $10,000 per intentional violation in an Attorney General action, and up to $10,000 per violation of a court order. Consumers may recover actual damages and attorney's fees. Boards may fine up to $1,500 per violation, suspend or revoke a license.
Conversational AI services must say they are AI when a person could think they are human, keep a suicide and self-harm protocol, and add protections for minors. They may not claim to be designed to provide professional mental or behavioral health care.
Signed
Effective
The AI sections apply from July 1, 2027. We have not confirmed the act's general effective date on an official source.
Covers addiction counselors?
Unclear. The AI sections do not mention substance use. They bar claims of providing “professional mental or behavioral health care” but do not say whether that includes addiction counseling.
Peer-support exemption
Not addressed.
Enforced by
Nebraska Attorney General, for the state or for any harmed person. No private right of action.
Penalties
At least $1,000 per violation, capped at $500,000 per operator, plus damages, costs and attorney's fees.
AB 406 (Statutes of Nevada 2025, ch. 283) AI in mental and behavioral health care
AI may not be offered or described as providing mental or behavioral health care, or called a “therapist” or “counselor.” Licensed providers may not use AI in direct patient care. They may use it for administrative work such as scheduling, billing and organizing session notes, and must independently check the accuracy of what AI produces for billing records and session notes.
Signed
Effective
. Parts took effect June 5, 2025, for regulations and preparation; July 1, 2025, for all other purposes.
Covers addiction counselors?
Yes. Covered care includes counseling people with alcohol and other substance use disorders, and covered providers include licensed or certified alcohol and drug counselors.
Peer-support exemption
Not addressed. Its exemptions cover self-help materials and administrative AI tools, not peer support.
Enforced by
Nevada Division of Public and Behavioral Health; licensing boards discipline licensed providers.
Penalties
Up to $15,000 per violation. A licensed provider's use of AI in direct care is unprofessional conduct.
Note
The act has no short title; the heading is our description.
HB 143 (Chapter 270, Laws of 2025) AI chatbot messages that endanger a child
It is a crime for the operator of an AI chat service to knowingly send a child messages meant to get the child to engage right away in sexually explicit conduct, “the illegal use of drugs or alcohol,” self-harm or suicide, or violence against another person.
Signed
Effective
Covers addiction counselors?
No. It applies to operators of AI chat services and does not regulate counselors' own practice. It names “the illegal use of drugs or alcohol” as one of the acts a chatbot operator may not steer a child toward.
Peer-support exemption
Not addressed.
Enforced by
New Hampshire Attorney General, who has the sole right to sue, after 90 days' written notice to fix the problem.
Penalties
A misdemeanor, plus civil damages of at least $1,000 per violation.
Note
The heading is our description. The law's official title says it creates a private right of action, but its text gives the Attorney General the sole right to sue. It is a criminal and civil liability law for AI chat services and has no crisis-protocol or disclosure duty.
S.3008-C / A.3008-C, Part U (Laws of 2025, ch. 58) Artificial Intelligence Companion Models (General Business Law Article 47)
An AI companion must have a protocol to detect suicidal thoughts or self-harm and refer users to 988 or other crisis services. It must tell users they are not talking with a human at the start of a conversation and at least every three hours if it continues.
Signed
Effective
. The law says it takes effect 180 days after it became law. November 5, 2025 is our count from the May 9 signing; the law does not print the date.
Covers addiction counselors?
No. It applies to whoever runs an AI companion, whatever their profession, and does not regulate counselors' own practice. The text does not mention substance use or addiction.
Peer-support exemption
Not addressed.
Enforced by
New York Attorney General.
Penalties
Up to $15,000 per day, paid into the state suicide prevention fund.
S.9008-C / A.10008-C, Part Y (Laws of 2026, ch. 58) Safe by Design Act
Online platforms with a built-in AI companion must check users' ages and turn the companion off by default for anyone under 18. A parent may turn it on.
Signed
Effective
Covers addiction counselors?
No. It sets default settings for minors on online platforms and does not regulate counselors or treatment.
An AI companion must say it is not human when a person could be misled, and must keep and publish a protocol that refers users with suicidal or self-harm thoughts to the 988 lifeline. For users it knows or has reason to believe are minors, it must remind them at least every three hours to take a break and must use reasonable measures to block sexually explicit content.
Signed
Effective
Covers addiction counselors?
No. It applies to whoever runs an AI companion, whatever their profession, and does not regulate counselors' own practice. The text has no provisions on substance use.
Peer-support exemption
Not addressed.
Enforced by
People who are harmed can sue (a private right of action). No state agency is named.
Penalties
The greater of actual damages or $1,000 per violation, plus a court order; the court may award attorney fees.
Note
The signing date comes from the Legislature's bill history; the signature block on the enrolled bill is blank.
H 7349 / S 2197 (Public Laws 2026, chs. 373 and 374) Oversight of Artificial Intelligence Technology in Mental Health Care Act
No one may provide, advertise or offer therapy in Rhode Island, including through AI, unless a licensed professional provides it. Licensed professionals may not let AI make independent therapy decisions or set treatment plans.
Signed
Effective
Covers addiction counselors?
Unclear. Its ban on offering therapy applies to anyone, and its limits on AI use apply to people who hold a Rhode Island license, credential or certification to provide therapy. We have not confirmed whether Rhode Island's addiction-counselor credentials count. It names substance use in its peer-support exemption.
Peer-support exemption
Yes. Peer support is exempt: encouragement and guidance from people with lived experience of “recovery from substance use or gambling,” given “without clinical intervention.” The definition describes services from people; the law does not say whether an AI tool can count as peer support.
Enforced by
Executive Office of Health and Human Services (EOHHS), which may investigate and make rules.
Penalties
No fine specific to the therapy ban; the confidentiality-breach penalties of R.I. Gen. Laws § 5-37.3-9 apply.
S 2195 / H 7350 (Public Laws 2026, chs. 376 and 375) Artificial Intelligence Companion Models
An AI companion must have a protocol for suicidal thoughts, self-harm and harm to others that refers users to crisis services. It must say it is not human at the start of a conversation and at least every three hours.
Signed
Effective
. Annual reports to the Attorney General begin July 1, 2027.
Covers addiction counselors?
No. It applies to whoever runs an AI companion, whatever their profession, and does not regulate counselors' own practice. The text does not mention substance use or addiction.
Peer-support exemption
Not addressed.
Enforced by
Rhode Island Attorney General.
Penalties
Up to $15,000 per day, directed to suicide prevention programs.
SB 1580 / HB 1470 (Public Chapter 647) AI presented as a mental health professional
A person who builds or deploys an AI system may not advertise or present it to the public as being, or able to act as, a qualified mental health professional.
Signed
Effective
Covers addiction counselors?
Unclear. It uses Tennessee's defined term “qualified mental health professional.” We have not yet confirmed whether that definition includes addiction counselors.
Peer-support exemption
Not addressed.
Enforced by
Enforced under the Tennessee Consumer Protection Act; the law does not name an agency.
Penalties
$5,000 per violation.
Note
The heading is our description. The official caption is “AN ACT to amend Tennessee Code Annotated, Title 33; Title 47 and Title 63, relative to mental health.”
A mental health chatbot must say it is AI, not a human, before first use, when a user returns after 7 days away, and whenever asked. It may not sell or share users' chat input. It may not sell or share their health information either, except in limited cases: a provider's request with the user's consent, the user's own health plan at the user's request, or a contractor bound by HIPAA-level safeguards. Ads inside a chat must be labeled.
Signed
Effective
Covers addiction counselors?
No. It applies to whoever supplies a mental health chatbot, whatever their profession, and does not regulate counselors' own practice. The text does not mention substance use or addiction.
Peer-support exemption
Not addressed.
Enforced by
Utah Division of Consumer Protection. The law does not provide for lawsuits by individuals.
Penalties
Up to $2,500 per violation, and up to $5,000 per violation of an administrative or court order.
Note
Two sections were amended in 2026, effective May 6, 2026; the fine amounts did not change.
H.816 (Act 156 of 2026) An act relating to regulating the use of artificial intelligence in the provision of mental health services
A corporation or other entity may not provide, advertise or offer mental health services to the public, including through AI, unless a mental health professional provides them or they are part of an approved research study. Professionals may use HIPAA-compliant AI tools if they review and approve the services. Prohibited AI use is unprofessional conduct for licensed professionals.
Signed
Effective
Covers addiction counselors?
Yes. Its definition of mental health professional includes alcohol and drug abuse counselors and peer recovery support specialists. Therapeutic communication includes interactions that “provide ongoing recovery support.”
Peer-support exemption
No. Peer support providers and peer recovery support specialists are listed as mental health professionals, not exempted.
Enforced by
The Vermont Attorney General and State's Attorneys, under the Consumer Protection Act, and private parties can sue. The Office of Professional Regulation or the Board of Medical Practice disciplines licensed professionals.
Penalties
Up to $10,000 per unfair or deceptive act in a state action, and up to $10,000 per violation of a court order. A consumer may recover damages or what they paid, attorney's fees, and exemplary damages up to three times what they paid.
An AI companion chatbot must say it is AI at the start of an interaction and at least every three hours, with hourly notices and rules against manipulation for minors. It must keep a suicide, self-harm and eating-disorder protocol that refers users to crisis services, and publicly report the protocol and the number of crisis referrals.
Signed
Effective
Covers addiction counselors?
No. It applies to whoever runs an AI companion chatbot, whatever their profession, and does not regulate counselors' own practice. The text does not mention substance use.
Peer-support exemption
Not addressed.
Enforced by
The Washington Attorney General, under the Consumer Protection Act. A person injured in their business or property can also sue under that Act.
Penalties
Up to $7,500 per violation in a state action, plus $5,000 more when the practice targets or affects people based on characteristics such as age or disability, and up to $125,000 for violating a court order. A person injured in their business or property may recover actual damages, costs and attorney's fees; a court may raise damages to up to three times actual damages, with the increase capped at $25,000.
Bills that have passed a state legislature but are not law yet. They are not included in the state list above.
California SB 903 Mental health professionals: artificial intelligence
It would limit AI in psychotherapy to administrative or supplementary support, require the patient's consent before AI records or transcribes therapy sessions or triage or screening, bar AI from doing triage or screening without a licensed professional's review and approval, and bar advertising psychotherapy delivered by companion chatbots. It defines psychotherapy as diagnosing or treating a mental health or substance use disorder, and exempts peer support, including from people in recovery from substance use disorders. Its list of licensed professionals does not name alcohol and drug counselors.
Status
Passed by the Legislature and presented to the Governor on September 9, 2026. No signature or veto was recorded as of September 28, 2026.
Deadline
. The last day for the Governor to sign or veto it, per the official California Senate calendar. If the Governor does not act by then, it becomes law without a signature (California Constitution, Article IV, section 10(b)(2)).
Call 911 and give naloxone (Narcan) now if you have it. If they don't wake up in 2 to 3 minutes, give another dose. Keep giving a dose every 2 to 3 minutes until they wake up. Stay with them until help arrives.