California SB 243 is a state law regulating AI companion chatbot platforms, authored by Sen. Padilla and signed October 13, 2025. It is the first US legislation directed specifically at AI companion and relationship products, imposing disclosure, suicide-prevention, minor-protection, reporting, and private-enforcement requirements on platforms that sustain ongoing relationships with users.
Status and timeline
The bill was signed into law on October 13, 2025, with a January 1, 2026 effective date for its core provisions. Its reporting requirements take effect July 1, 2027.
Scope and definitions
SB 243 applies to "companion chatbot platforms" — AI systems that sustain relationships with users across multiple interactions. Its scope is capability-based: coverage turns on what a system is capable of doing rather than on observed conduct.
The statute defines a "companion chatbot" as an AI system with a natural language interface that provides adaptive, human-like responses to user inputs and is capable of meeting a user's social needs, including by exhibiting anthropomorphic features and sustaining a relationship across multiple interactions (Source: gunder.com). Compliance obligations fall on the "operator" — the person or entity making a companion chatbot platform available to a user in California — even where the chatbot is built on a third-party vendor's AI model; the operator cannot delegate its compliance obligations to the underlying model provider (Source: gunder.com).
The law expressly excludes bots used only for customer service, business operations, productivity, internal research, or technical assistance; video-game bots limited to game-related replies that cannot discuss mental health, self-harm, or sexually explicit conduct; and standalone voice-command consumer devices (speakers/voice assistants) that do not sustain a relationship across interactions or generate outputs likely to elicit emotional responses (Source: gunder.com). Legal analyses note that coverage can nonetheless reach customer-service, virtual-assistant, wellness-coaching, and academic-support chatbots when they recall prior conversations, adapt to a specific user, or build rapport across sessions — the operative test being whether the tool is "capable of meeting a user's social needs" (Source: gunder.com).
Key provisions
The statute requires covered platforms to:
- Disclose that chatbots are AI, where a reasonable person might otherwise believe they are human.
- Maintain published suicide prevention protocols.
- Provide enhanced protections for known minors.
- File annual reports with the Office of Suicide Prevention.
- Allow a private right of action for individuals injured by noncompliance.
Relative to the other West Coast companion-chatbot statutes, SB 243 is the most minimal: it is disclosure-focused and imposes minimal engagement-design requirements.
Disclosure obligations
For all users, the AI disclosure must be a "clear and conspicuous notification" wherever a reasonable person could be misled into thinking they are interacting with a human — an obligation that operates in addition to California's existing SB 1001 bot-disclosure law for commercial and electoral bots. Operators must also clearly and conspicuously disclose, on the application, browser, or other access format, that the companion chatbot may not be suitable for some minors; legal analyses read this as requiring an in-app banner, popup, or similarly prominent notice rather than a clause in standard terms (Source: gunder.com).
For users the operator knows are minors — with knowledge read as actual or constructive awareness that a user is under 18 — the operator must additionally provide a clear and conspicuous notification at least every three hours during a continuous interaction, reminding the user to take a break and that the chatbot is AI-generated. The requirement applies per individual minor user (Source: gunder.com).
Harm prevention protocols
Companion chatbots may not engage with users unless the operator implements and maintains safety protocols to prevent the production of suicidal-ideation, suicide, or self-harm content. Operators must publish details of these protocols on their websites, including notifications referring at-risk users to crisis service providers (suicide hotlines, crisis text lines) when a user expresses suicidal ideation or self-harm. For known minors, operators must also institute reasonable measures preventing the chatbot from producing sexually explicit visual material or directly stating that the minor should engage in sexually explicit conduct (Source: gunder.com).
Annual reporting (from July 1, 2027)
Annual reports to the California Department of Public Health's Office of Suicide Prevention must detail the number of crisis-service-provider referrals issued in the preceding calendar year, the protocols for detecting, removing, and responding to instances of suicidal ideation, and the protocols prohibiting chatbot responses about suicidal ideation or actions. Suicidal ideation must be measured using evidence-based methods, and reports may not include user identifiers or personal information (Source: gunder.com).
Enforcement
SB 243 creates a private right of action for any person who suffers "injury in fact" from a violation: injunctive relief, damages equal to the greater of actual damages or $1,000 per violation, and reasonable attorneys' fees and costs. The statute states that its duties are cumulative with operators' obligations under other applicable laws (Source: gunder.com).
Background
The statute responds to teen chatbot deaths cited in the AI LEAD Act's findings, most prominently the facts of Garcia v. Character Technologies (Sewell Setzer III, Feb. 2024) and Raine v. OpenAI (Adam Raine, April 2025). Character.AI and Replika are the paradigm regulated products, and the underlying concern is treated at the concept level in AI Mental Health and Psychological Harm, the case the statute instantiates.
The statute's structure tracks the APA's 2025 AI health advisories — disclosure, crisis-escalation, and default-on youth protection. Anthropic's "Protecting the Wellbeing of Users" describes an engineering analog to the statutory requirements (Source: anthropic.com). Child safety is one of the carve-outs protected from preemption in EO 14365.
West Coast companion laws
Oregon and Washington enacted companion chatbot laws in 2026, both effective January 1, 2027 — one year after SB 243's January 1, 2026 effective date (Source: The Rest of the West: Oregon and Washington Build on California Chatbot Law).
| State | Law | Key distinction |
|---|---|---|
| California | SB 243 | Capability-based scope; disclosure-focused; minimal engagement design requirements |
| Oregon | SB 1546 | Behavior-based scope (narrower); excludes patient care services; private right of action |
| Washington | HB 2225 | Capability-based scope; most prescriptive (disclosure every 3 hours); explicit engagement design restrictions |
Washington HB 2225 is the most expansive of the three: it explicitly prohibits chatbot design techniques likely to create compulsive use or unhealthy attachment in minors. California's SB 243 is the most minimal. Oregon's SB 1546 occupies a middle ground, with a narrower behavior-based scope definition but similar requirements. Companies operating chatbots across all three states face overlapping but non-identical requirements, and may default to the most restrictive standard, Washington's, across all three.
Relationships
- depends-on: AI Mental Health and Psychological Harm — the concept-level case the statute instantiates.
- related: Character.AI, Replika — the paradigm regulated products.
- related: Garcia v. Character Technologies, Raine v. OpenAI — fact patterns in legislative findings.
- related: APA AI Health Advisories — framework the statute partially codifies.
- related: EO 14365 — child safety carve-out from preemption.