US states split on how to regulate AI mental health chatbots

Oregon enacted Senate Bill 1546 in 2026, requiring covered AI companion operators to maintain suicide and self-harm protocols and allowing users who suffer ascertainable harm to seek damages or an injunction. Other US states have drawn their boundaries elsewhere, regulating therapy, professional claims or general-purpose conversational systems rather than one common category of mental health chatbot.

The result is not a simple scale from permissive to strict. A service's legal position can turn on whether it presents itself as therapy, simulates an ongoing relationship, serves minors or is used by a licensed professional. The same product may therefore face different duties as its marketing, features and location change.

Illinois and Maine draw a line around therapy

Illinois Public Act 104-0054 bars an individual or company from offering therapy or psychotherapy in the state unless the service is conducted by a professional licensed in Illinois. A licensed professional may use AI for administrative or supplementary support while retaining responsibility, but may not let the system make independent therapeutic decisions, communicate directly with clients as therapy or create unreviewed treatment plans.

The Illinois law does not prohibit every conversational tool that discusses emotions. Its definitions exclude religious counseling and peer support, and its operative provisions turn on whether a service provides or advertises therapy or psychotherapy. That makes product claims and the nature of the interaction part of the coverage question.

Maine Public Law Chapter 687 likewise requires licensed professionals to remain responsible for AI-assisted therapy and bars AI from making independent therapeutic decisions or directly conducting therapeutic communication. When AI records or transcribes a session for supplementary support, the client must receive written information about the tool's purpose and the handling of session data, then provide consent.

Maine also says a professional cannot refuse therapy solely because a client declines that AI use. The statute applies its controls across several licensed professions, making professional responsibility and consent its main regulatory levers.

Oregon and Nebraska regulate broader conversational systems

Oregon's law is built around an “artificial intelligence companion,” defined through features such as sustained human-like interaction, memory of previous exchanges and personalized engagement. Its official summary says covered operators must disclose that users are interacting with artificial output, maintain protocols for suicide and self-harm, add protections for minors and publish annual reports on crisis referrals.

The law also creates a private enforcement route. A user who suffers an ascertainable loss or injury from a violation may bring a claim for damages and an injunction. This differs from a licensing rule aimed only at therapists because it places duties directly on operators of covered companion products.

Nebraska uses the wider category of a publicly available conversational AI service that primarily simulates human conversation through text, images or audio. Nebraska Revised Statute 86-1805 requires operators to adopt a protocol that responds to suicide or self-harm prompts and makes reasonable efforts to refer users to appropriate crisis services.

The Nebraska act becomes operative on July 1, 2027. Other sections require disclosure when a reasonable person could mistake the system for a human and set protections for minor account holders. Its definition excludes narrow-topic tools, business-only systems, voice assistants and customer-service software, showing that not every chatbot falls within the law.

Tennessee regulates professional claims

Tennessee Public Chapter 647, effective July 1, 2026, prohibits developers and deployers from advertising or representing an AI system as a qualified mental health professional. A violation is treated as an unfair or deceptive practice under the US state's consumer-protection law.

This is narrower than banning all emotionally supportive conversation. The trigger is a public claim that the system is, or can act as, a qualified mental health professional. A service may therefore avoid this particular provision while still coming within another state's therapy or companion-AI rules.

The patchwork starts with definitions

The five states regulate at least three different relationships. Illinois and Maine govern therapy and the role of licensed professionals. Oregon and Nebraska impose duties on operators of defined companion or conversational services, while Tennessee targets claims of professional qualification.

Those choices produce different compliance questions. Therapy laws ask whether an interaction amounts to therapeutic communication and whether a licensed person retains control. Companion laws ask how the system is designed, what it remembers and whether it can identify crisis-related prompts. Tennessee asks what the operator tells the public about the system's status.

This report does not extend those state-specific rules to users elsewhere. It also does not assess a product's clinical quality or recommend any mental health service. Regulatory status depends on the service, the jurisdiction and the facts of a particular use.

What remains outside this comparison

The source report also made claims about Taiwan's laws and agency activity. APPI News could not verify those current-status claims through an English-language primary source, so they are not repeated here. The omission should not be read as a finding that Taiwan has no applicable law or oversight.

State rules also continue to develop through implementation and litigation. The enacted texts establish duties and remedies, but future agency guidance and court decisions may determine which products fall within contested terms. A comparison based on statutory text cannot resolve those product-specific questions in advance.