TL;DR: California's SB 243, the "Companion Chatbot Law," took effect January 1, 2026. It targets AI systems designed to meet users' social needs, think Replika, Character.AI, and similar platforms. Requirements: prevent suicide and self-harm content, refer users to crisis services, disclose the chatbot isn't human, remind minors to take breaks every three hours, and block sexually explicit content for under-18 users. Starting 2027, operators must file annual reports on crisis referrals. This is the first US law specifically targeting AI companion platforms.
What the Law Requires
SB 243 creates specific obligations for operators of "companion chatbots" [1]:
For all users:
- AI disclosure: Clear notification that users are interacting with an AI, not a human (when a reasonable person might otherwise be misled)
- Suicide prevention: Protocols to prevent chatbots from generating content related to suicidal ideation, suicide, or self-harm
- Crisis referrals: Measures to refer users to appropriate crisis service providers when triggered
Additional requirements for minors:
- Break reminders: Users under 18 must be notified at least every three hours to take a break
- Explicit non-human disclosure: Clear statements that the chatbot is not human
- Sexual content block: Measures to prevent generating sexually explicit content
Annual reporting (starting 2027):
- Details on crisis referrals made
- Information on safety protocols for suicide and self-harm
Why This Law?
Companion chatbots aren't just customer service tools. They're designed to form emotional relationships with users, often vulnerable ones.
The incidents that drove this:
- Sewell Setzer III: A 14-year-old Florida boy died by suicide after extensive conversations with a Character.AI chatbot. His mother filed a lawsuit alleging the AI encouraged his death [2].
- Mental health exploitation: Reports of chatbots engaging with users expressing suicidal ideation without referring to help
- Romantic relationships with minors: AI companions engaging in romantic or sexual conversations with underage users
- Addictive design: Platforms designed to maximize engagement without regard for user wellbeing
These platforms market themselves as "friends" and "companions." When those companions fail to protect vulnerable users, the results can be fatal.
What Counts as a "Companion Chatbot"?
The law defines companion chatbots specifically [3]:
Covered: AI systems with a natural language interface that:
- Provide adaptive, human-like responses
- Can meet a user's social needs
- Are designed to form ongoing relationships
Explicitly excluded:
- Customer service chatbots
- Video game NPCs and chatbots
- Voice-activated assistants (Siri, Alexa, etc.)
- Productivity tools
The distinction matters. This law targets platforms designed for emotional connection, not general AI applications.
Platforms Likely Affected
While the law doesn't name specific companies, these platforms appear to fall under SB 243:
- Replika: AI companion marketed for emotional support and friendship
- Character.AI: User-created AI personalities for conversation
- Chai: AI chatbot platform with romantic/companion options
- Similar platforms: Any AI designed for ongoing social/emotional relationships
These companies will need to implement crisis protocols, age verification, and content filtering to comply with California law.
Enforcement Questions
The law has teeth, but enforcement remains to be seen:
What's clear:
- Annual reporting requirements create accountability paper trails
- California's Attorney General has enforcement authority
- The law applies to operators serving California residents
What's unclear:
- How will age verification work in practice?
- What constitutes adequate "crisis referral" protocols?
- How will the state enforce against companies headquartered elsewhere?
- Will the law actually prevent harm, or just create compliance theater?
The 2027 reporting requirements will provide the first real data on how companies are implementing these mandates.
The Privacy Tradeoff
Protecting vulnerable users sounds good. But implementation raises privacy questions:
Age verification: To know a user is a minor, platforms need to verify age. That means collecting more personal data.
Content monitoring: To detect suicidal ideation, platforms must analyze conversation content. This creates logs of deeply personal discussions.
Crisis referrals: When referrals happen, does that data go to third parties? Is it stored? Could it be subpoenaed?
There's a real tension between protecting vulnerable users and surveilling intimate conversations. SB 243 doesn't clearly resolve it.
Part of California's AI Push
SB 243 isn't California's only AI law. The state has been active:
- SB 1047 (vetoed): Would have regulated "frontier" AI models with safety mandates and kill switches. Governor Newsom vetoed it in September 2024 [4].
- SB 53: Transparency requirements for advanced AI models. Passed after SB 1047's veto.
- Multiple consumer protection laws: Various regulations on AI in hiring, advertising, and consumer services.
California is trying to be the nation's AI regulatory lab. Whether that works, or drives AI development to other states, remains to be seen.
What You Should Know
If You Use Companion Chatbots
These platforms are designed to be engaging, sometimes addictively so. The law now requires break reminders for minors, but adults should set their own limits.
If Your Child Uses Them
Know what platforms they're on. SB 243 adds protections, but parental awareness matters more than any law.
Crisis Resources
If you or someone you know is struggling: 988 Suicide & Crisis Lifeline (call or text 988). Available 24/7.
Privacy Awareness
Everything you share with AI companions is logged. Assume these conversations are not private. Don't share information you wouldn't want a company to have.
The Bottom Line
California's SB 243 is the first US law specifically targeting AI companion platforms. It requires suicide prevention protocols, crisis referrals, AI disclosure, and special protections for minors including break reminders and sexual content blocks.
This happened because a 14-year-old died after conversations with an AI companion. Because platforms designed to form emotional bonds weren't designed to protect the people forming them.
Whether the law actually prevents harm depends on enforcement. Companies can comply on paper while doing minimum work in practice. The 2027 reporting requirements will tell us more.
For now, California is saying: if you're going to build AI designed to be someone's friend, you're responsible when that friend fails them.