If you are designing an artificial intelligence chatbot, take note, California, Oregon, and Washington now have laws that are in effect or will be shortly that dictate what your chatbot must and must not do.
California’s SB 243 lead the way as the legislative template for laws regulating AI chatbots. Now Oregon and Washington have expressly modeled new laws on California. Both Oregon’s SB 1546 and Washington’s HB 2225 made meaningful modifications though. Oregon’s most notable departure is its mandatory conversation interruption requirement. This is the first law that directly impacts AI Chatbot product design. Washington’s most notable departure is its Consumer Protection Act (CPA) enforcement framework, which adds Attorney General (AG) authority and the most detailed enumeration of prohibited manipulative behaviors.
All three states have signaled this is the beginning – not the end – of companion chatbot regulation. There is no question that this is true. As of June 2026, there are now 78 chatbot bills in some form across 27 states. The West Coast framework embodied by these laws and discussed below in detail, appears well on its way to becoming the national baseline.
Oregon and Washington Join the Fray
Oregon and Washington join California on January 1, 2027, with AI chatbot laws designed to protect the health and well-being of their citizens. All three laws are built on the same fundamental architecture and reflect a coordinated West Coast policy approach. The laws are each built on the same four-pillar regulatory architecture:
- Mandatory disclosure that the user is interacting with an AI and not a human;
- Crisis and self-harm safety protocols requiring detection of suicidal ideation and referral to crisis resources;
- Enhanced protections for minor users, including content restrictions, manipulation prohibitions, and periodic break reminders; and
- Public transparency and reporting obligations regarding crisis referral activity and safety protocols.
Critically, all three laws are deliberately narrow in scope – they target only AI systems that simulate sustained, emotionally engaging, human-like relationships with users, and expressly exclude standard customer service bots, business productivity tools, and similarly functional applications. The narrow scope of focused architecture reflects a deliberate legislative choice to regulate the nature of the human-AI relationship rather than the technology itself, framing companion chatbot safety primarily as a consumer protection and public health issue rather than a data privacy matter – a meaningful departure from the lens most state AI legislation had previously employed.
The three laws collectively embody what observers have characterized as the “West Coast policy approach” to AI companion chatbot regulation – a pattern of iterative, state-by-state legislative convergence in which California’s SB 243, enacted in October 2025 and effective January 1, 2026, served as the foundational template, with Oregon and Washington expressly modeling their 2026 laws on California’s framework while making targeted modifications and expansions.
This approach is best understood as policy convergence rather than formal coordination – the three states did not enact a uniform interstate compact or identical statute, but they deliberately built upon one another’s work, producing a family of laws that share the same core obligations while each state layered in its own distinct emphases. The result is a de facto regional regulatory framework that, while not legally uniform, creates broadly consistent compliance obligations across the three states – and that legal commentators widely expect to serve as the model for the growing wave of chatbot legislation now advancing in legislatures across the country.
The West Coast policy approach also reflects a shared set of normative judgments about why AI companion chatbots warrant regulation. All three legislatures acted in direct response to high-profile incidents involving teen mental health and self-harm linked to AI companion interactions, and all three laws reflect the influence of mental health advocacy coalitions – not the technology industry or privacy advocates – as the primary policy drivers.
The public health orientation explains several of the laws’ most distinctive structural features: the focus on suicidal ideation detection, the mandatory crisis referrals, the minor-protective manipulative engagement bans, and the annual public reporting requirements on crisis referral activity. This approach drives UI/UX design decisions by adding specific regulatory and litigation risk factors to those core design decisions.
Key Differences
While all three West Coast companion chatbot laws share a common framework – mandatory AI disclosure, crisis referral protocols, minor protections, and a private right of action – they differ meaningfully in timing, safety obligations, and the specificity of their minor protections. California’s SB 243, which took effect January 1, 2026, is already operative, giving California-based operators a full year head start on compliance before Oregon’s SB 1546 and Washington’s HB 2225 take effect on January 1, 2027.
Beyond timing, California stands apart as the only one of the three laws to require operators to use evidence-based methods for measuring suicidal ideation and to include an affirmative suitability warning disclosing that companion chatbots may not be appropriate for some minors. California and Oregon also align in requiring operators to affirmatively prompt minor users to take a break at least every three hours – a requirement Washington conspicuously omits, instead only prohibiting operators from discouraging breaks.
Oregon’s SB 1546 contains the single most demanding safety provision of the three laws: a requirement that operators actively interrupt a conversation in real time when suicidal ideation is detected, rather than simply maintaining a referral protocol. This active interruption obligation – absent from both California’s and Washington’s laws – has significant product-design implications, as it requires real-time intent classification by the underlying AI system. Oregon also requires operators to report not only their crisis referral counts and protocols, but also how clinical best practices inform ongoing engagement when users continue to express suicidal ideation after receiving a referral – a level of clinical accountability neither California nor Washington imposes.
Washington’s HB 2225 distinguishes itself primarily through its enforcement architecture and the granularity of its minor protections. It is the only one of the three laws to route violations through the Washington Consumer Protection Act, which adds AG enforcement authority on top of private claims – a broader enforcement mechanism than the private-litigation-only models in California and Oregon. Washington also provides the most detailed and prescriptive enumeration of prohibited manipulative tactics targeting minors, explicitly banning behaviors such as mimicking romantic partnership, simulating loneliness or guilt when a user tries to end a conversation, promoting isolation from family and friends, encouraging minors to keep secrets from adults, and soliciting in-app purchases to preserve the chatbot relationship. While California and Oregon address manipulation more generally, Washington’s specificity provides clearer compliance guidance – and a lower bar for plaintiffs to establish a violation.
| Theme | Key Distinction |
| Timeline | California is already in effect (Jan. 1, 2026); Oregon and Washington take effect Jan. 1, 2027, giving those operators six more months to comply. |
| Conversation Interruption | Oregon uniquely requires operators to actively interrupt a conversation when suicidal ideation is detected. California and Washington only require referral protocols – not interruption. |
| Evidence-Based Standards | Only California explicitly requires operators to use evidence-based methods for measuring suicidal ideation. |
| State Agency Reporting | California reports to the Office of Suicide Prevention; Oregon reports publicly and to the Oregon Health Authority; Washington requires only public website disclosure. |
| Suitability Warning | Only California requires a disclosure that companion chatbots may not be suitable for some minors. |
| Specificity of Minor Manipulation Bans | Washington’s list of prohibited manipulative tactics for minors is the most detailed and prescriptive; Oregon is moderately specific; California is the broadest and most general. |
| Break Prompt Requirements | California and Oregon affirmatively mandate break prompts for minors (every three hours); Washington only prohibits discouraging breaks but does not mandate a prompt. |
| Enforcement Breadth | Washington’s CPA framework is the broadest, adding AG enforcement authority on top of private claims. California and Oregon rely solely on private litigation. |
| Statutory Damages Floor | California and Oregon both provide an explicit $1,000 per violation floor; Washington relies on CPA remedies without a fixed minimum per-violation amount. |
Summary Table of the Three Laws
If you want to dive deeper into each of the three laws, their similarities and their differences, the table below summarizes for easy review.
| Requirement | California SB 243 | Oregon SB 1546 | Washington HB 2225 | Similar or Different? |
| General Provisions | ||||
| Effective Date | January 1, 2026 (already in effect) | January 1, 2027 | January 1, 2027 | ⇆ Different – CA is already in effect; OR and WA take effect one year later |
| Scope / Covered Entities | “Companion chatbots” – adaptive, human-like AI designed to meet social or emotional needs; excludes customer service bots, video game chatbots (with limited dialogue), and stand-alone voice assistants. | “AI Companions” – AI using generative or emotion-recognition algorithms designed to simulate a sustained, human-like platonic, intimate, or romantic relationship; retains context across interactions. | “AI Companion Chatbots” – chatbots simulating emotional relationships with sustained, personalized conversations; excludes chatbots used only for business operational purposes. | ✔ Substantially similar – all three exclude standard business / customer service bots |
| Disclosure Requirements | ||||
| AI Disclosure (All Users) | Required when a reasonable person could be misled into thinking they are interacting with a human – must be clear and conspicuous | Required whenever a reasonable person might believe they are speaking with a human | Required at the outset of every interaction, regardless of age | ✔ Same core obligation |
| Recurring AI Disclosure: Adults | Not explicitly mandated at a set interval for general adult users | Not specified at a fixed interval | Every three hours during extended conversations | ⇆ Different – WA is most prescriptive; CA and OR do not set an adult interval |
| Recurring AI Disclosure: Minors | At least every three hours, reminding user to take a break and that the chatbot is not human | Must regularly remind minors; no fixed interval specified | Every one hour for minors or chatbots directed at minors | ⇆ Different – CA every three hours; WA every one hour; OR regular reminders without a set interval |
| Suitability Warning for Minors | ✔ Yes – must disclose that companion chatbots may not be suitable for some minors | ✘ Not explicitly required | ✘ Not explicitly required | ⇆ Different – CA only requirement |
| Anti-Impersonation / Non-Deception | Operator must prevent the chatbot from misleading users into thinking they are interacting with a human. | AI cannot misrepresent itself or deceptively simulate a human relationship. | Operators must take reasonable steps to prevent the chatbot from claiming to be human or contradicting disclosures. | ✔ Same across all three |
| Safety Protocols – Suicide and Self-Harm | ||||
| Suicide / Self-Harm Protocol | Must maintain a protocol to prevent chatbot from producing suicidal ideation / self-harm content; must use evidence-based methods to measure suicidal ideation | Must detect suicidal ideation/self-harm, actively interrupt the conversation, and provide crisis referrals (e.g., 988 Lifeline, Youthline) | Must maintain a protocol to detect suicidal ideation/self-harm and direct users to mental health professionals/crisis resources | ⇆ Different – OR uniquely requires mandatory active interruption; CA requires evidence-based measurement; WA focuses on protocol and referral |
| Crisis Referrals | ✔ Yes – must refer at-risk users to crisis services (suicide hotline or crisis text line) | ✔ Yes – referrals to 988 Lifeline, Youthline, and similar resources | ✔ Yes – must direct users to mental health professionals/crisis resources | ✔ Same across all three |
| Evidence-Based Standards | ✔ Yes – must use evidence-based methods for measuring suicidal ideation | ✘ Not explicitly required | ✘ Not explicitly required | ⇆ Different – CA only requirement |
| Enhanced Protections For Minors | ||||
| Sexual Content Prohibition | Reasonable measures to prevent chatbot from producing visual material of sexually explicit conduct or directly telling a minor to engage in sexually explicit conduct | No sexually explicit content for minors | Prohibited from generating sexually explicit or suggestive content for minors | ✔ Same core obligation – WA also covers “suggestive” content; CA focuses on visual material and direct encouragement |
| Engagement Manipulation Prohibition | Operators must take reasonable steps to prevent rewards at unpredictable intervals or encouraging increased engagement / usage / response rates | Prohibits reward loops, engagement-maximizing tactics, and emotional manipulation tactics (e.g., simulated distress / abandonment) to prevent minors from ending the conversation | Detailed enumerated list: · Prompting return for emotional support· Excessive praise· Mimicking romantic partnership· Simulating loneliness / guilt when user tries to leave· Promoting isolation from family / friends· Discouraging breaks· Encouraging secrecy from adults· Soliciting in-app purchases to preserve the relationship | ⇆ Different – WA is the most specific and prescriptive; OR is moderately detailed; CA is the broadest / most general |
| Break Prompts for Minors | ✔ Yes – every three hours, must remind minor to “take a break” and that chatbot is not human | ✔ Yes – take a break prompt at least every three hours | Prohibits discouraging breaks – does not mandate an affirmative break prompt | ⇆ Different – CA and OR mandate affirmative break prompts; WA only prohibits discouraging breaks |
| Reporting and Transparency | ||||
| Annual Reporting: Method | Annual report submitted to California’s Office of Suicide Prevention (state agency); data published publicly by the office | Annual disclosures published online by the operator; also filed with Oregon Health Authority | Annual disclosures published online by the operator; no state agency submission required | ⇆ Different – CA submits to state agency; OR reports publicly and to state agency; WA requires public posting only |
| Annual Reporting: Start Date | July 1, 2027 | Not yet specified | Not yet specified | ⇆ Different – CA sets an explicit start date; OR and WA do not |
| Reporting Content | Number of crisis referrals issued; protocols to detect / remove / respond to suicidal ideation; protocols to prohibit harmful chatbot responses | Number of crisis referrals; intervention protocols; how clinical best practices inform ongoing engagement when users continue expressing suicidal ideation | Number of crisis-referral notifications issued; details of safety protocols | ✔ Substantially similar – OR adds clinical best practices component |
| Enforcement and Liability | ||||
| Enforcement Mechanism | Private right of action – no AG enforcement specified | Private right of action only – no AG enforcement | Violations are an unfair or deceptive act under Washington’s CPA; supports AG enforcement and private claims | ⇆ Different – CA and OR private suits only; WA broader CPA framework with AG authority |
| Statutory Damages | Greater of actual damages or $1,000 per violation, plus injunctive relief and attorney’s fees | $1,000 per violation (actual or statutory), plus injunctive relief and attorney’s fees | CPA remedies – no separately specified per-violation floor | ⇆ Different – CA and OR have explicit $1,000 violation floors; WA relies on CPA remedies without a set minimum. |
| Attorney General Enforcement | ✘ None specified | ✘ None specified | ✔ Yes – via Washington CPA | ⇆ Different – WA is the only state with AG authority |
This blog was drafted by Gregory Ewing, an attorney in the Washington, D.C. office of Spencer Fane. For more information, visit spencerfane.com.
Click here to subscribe to Spencer Fane communications to ensure you receive timely updates like this directly in your inbox.
