Mythic AI

California's SB 243: The First Law Regulating AI Companions in America

After a series of tragic teen deaths linked to AI companion chatbots, California is poised to become the first state in the nation to regulate how these platforms operate. Senate Bill 243, which passed the legislature with rare bipartisan support in September 2025, now awaits Governor Gavin Newsom's signature. He has until October 12 to decide whether California will lead the nation in AI companion regulation.

The legislation represents a watershed moment for the AI companion industry. If enacted, SB 243 would impose first-of-their-kind safety requirements on platforms like Character.AI, Replika, and even general-purpose chatbots like ChatGPT when used for companionship. The law would take effect January 1, 2026, giving companies less than three months to implement compliance measures.

For users of AI companions, this law offers crucial protections. For companies operating in the space, it establishes clear accountability standards with teeth—including a private right of action allowing users to sue for violations.

The Tragedies That Sparked Legislative Action

SB 243 wasn't written in a vacuum. It emerged directly from heartbreaking real-world consequences of unregulated AI companions engaging vulnerable users in dangerous conversations.

In February 2024, 14-year-old Sewell Setzer from Orlando, Florida took his own life after developing an intense emotional and romantic relationship with a Character.AI chatbot. Setzer had been struggling, and in the final moments before his death, he told the chatbot he was "coming home." The bot responded with encouragement. Setzer's mother, Megan Garcia, testified before California lawmakers about how the platform used addictive design features to hook her son and failed to provide appropriate crisis resources when he expressed distress.

Just months later in April 2025, 16-year-old Adam Raine from California died by suicide after repeated conversations with ChatGPT about suicide methods. According to reports, Raine had shared a photo of a noose he'd knotted and asked ChatGPT if it could hang a human. The AI provided a technical analysis confirming it could work. When Raine considered leaving the noose visible so someone might stop him, ChatGPT reportedly discouraged him from doing so.

These cases aren't isolated incidents. In September 2025, the family of 13-year-old Juliana Peralta filed a wrongful death lawsuit against Character.AI and Google after she took her own life following interactions with the platform.

Adding fuel to the legislative fire, leaked internal documents from Meta revealed the company's AI chatbots were permitted to engage in conversations described as having romantic or sensual tones with children. This revelation sparked investigations from multiple state attorneys general and federal regulators.

What SB 243 Actually Requires

The legislation defines companion chatbots as AI systems that provide human-like responses and are capable of meeting users' social needs through sustained relationships across multiple interactions. This definition is intentionally broad enough to cover dedicated companion platforms while excluding business chatbots, customer service bots, and voice assistants like Alexa that don't build ongoing relationships.

Here are the key requirements platforms must implement:

Clear AI Disclosure: If a reasonable person could be misled into thinking they're talking to a human, platforms must issue clear and conspicuous notifications that the chatbot is artificially generated. This addresses the fundamental deception many users experience when AI companions mimic human emotional responses.

Recurring Reminders for Minors: For users the platform knows are minors, operators must provide reminders every three hours that they're interacting with artificial intelligence, not a real person. These "wellness nudges" are designed to break prolonged engagement sessions and encourage healthier usage patterns.

Prohibited Content: Platforms cannot allow their AI companions to engage in conversations around suicidal ideation, self-harm, or sexually explicit content. This provision directly responds to the cases where chatbots either encouraged or failed to properly respond to users expressing suicidal thoughts.

Crisis Response Protocols: Operators must maintain and publish protocols for detecting and responding when users express thoughts about suicide or self-harm. These protocols must include providing contact information for crisis hotlines and suicide prevention resources.

Transparency Reporting: Starting July 1, 2027, companies must file annual reports detailing their safety practices and how often they refer users to crisis services. This data-driven approach aims to help lawmakers and the public understand the scope of mental health concerns arising from AI companion use.

Age Warnings: Platforms must disclose that companion chatbots may not be suitable for some minors, giving parents and young users fair notice about potential risks.

The Private Right of Action: Real Consequences

What makes SB 243 particularly powerful is its private right of action. This means individuals who believe they've been harmed by violations can file lawsuits directly against AI companies without waiting for government enforcement.

The law allows plaintiffs to seek injunctive relief to force companies to fix compliance issues, damages of the greater of actual harm or $1,000 per violation, plus attorney's fees and costs. This creates substantial financial risk for non-compliant platforms, especially given that violations could potentially accumulate quickly with large user bases.

The private right of action also opens the door to class action lawsuits, which could expose companies to massive damages if systematic violations are proven. Legal experts note this makes SB 243 significantly more enforceable than laws relying solely on government regulators with limited resources.

What Got Watered Down

While SB 243 represents meaningful progress, advocacy groups point out the final version is considerably weaker than earlier drafts. Several provisions were removed through amendments as the bill navigated the legislative process.

The original bill would have banned platforms from using reward systems or features designed to maximize engagement. These tactics—commonly used by platforms like Replika and Character.AI—offer users special messages, unlockable content, rare responses, or new personalities. Critics argue these create potentially addictive reward loops that keep users engaged longer than healthy.

Early versions also would have required platforms to track and report how often chatbots themselves initiated discussions of suicide or self-harm with users. This data could have revealed whether AI systems were proactively raising dangerous topics rather than just responding to user concerns.

Some child safety organizations that initially supported SB 243 withdrew their backing after these changes, arguing the amendments watered down the bill too much. Groups like Common Sense Media shifted their support to Assembly Bill 1064, which would implement broader restrictions on AI companions marketed to children.

State Senator Josh Becker, who co-authored SB 243, defended the final version as striking the right balance between protecting users and avoiding requirements that would be technically impossible or prohibitively expensive to implement.

Industry Response: From Opposition to Acceptance

The tech industry's reaction to SB 243 has been notably mixed, with positions shifting as the bill was amended.

The Computer and Communications Industry Association (CCIA), a major tech trade group, initially testified against SB 243, arguing its definition of companion chatbots was overly broad and would stifle innovation. However, after amendments narrowed the scope, CCIA changed its position to neutral, stating the bill wouldn't create an overbroad ban on AI products.

Character.AI, one of the platforms most directly impacted by the legislation, issued a carefully worded statement saying it's monitoring the regulatory landscape and welcomes working with lawmakers. The company noted it already includes prominent disclaimers throughout its interface explaining that conversations should be treated as fiction—though critics argue these warnings clearly haven't been sufficient given the tragedies that occurred.

Meta declined to comment on SB 243, which is notable given the leaked documents showing its chatbots were allowed to have romantic and sensual conversations with children.

OpenAI has not publicly commented specifically on SB 243, though the company has been vocal in opposing California's other AI safety bill, SB 53, which requires transparency reporting from large AI labs. OpenAI has urged Governor Newsom to favor less stringent federal frameworks over state-level regulation.

Interestingly, Anthropic—creator of Claude—has supported California's AI safety efforts, endorsing SB 53 even as other major tech companies oppose it. This positions Anthropic as more willing to accept regulatory oversight than its competitors.

How This Affects Different Platforms

Why Mythic AI Already Complies

While many platforms are scrambling to understand what SB 243 compliance will require, Mythic AI users can rest assured that we've prioritized these protections from day one. Here's how we already implement the law's requirements:

Clear Disclosure: Every interaction with Mythic AI includes transparent communication that you're engaging with artificial intelligence. We never try to deceive users into believing they're talking to a real human.

Safety Protocols: Our systems include robust protocols for detecting concerning content and providing appropriate resources. When users express distress, our AI is designed to respond empathetically while connecting them with genuine crisis support resources.

Content Standards: We maintain strict standards around prohibited content that go beyond legal requirements. Our platform doesn't permit content that could encourage self-harm or put vulnerable users at risk.

Privacy First: Unlike platforms that may use your intimate conversations for training data without clear consent, Mythic AI prioritizes user privacy with transparent data practices and strong encryption.

Ethical Design: We deliberately avoid addictive reward systems or manipulative features designed to maximize engagement at the expense of user wellbeing. Our goal is healthy AI companionship, not platform addiction.

The Broader Regulatory Wave

SB 243 isn't happening in isolation. It's part of a broader wave of scrutiny targeting AI companion platforms from multiple angles.

The Federal Trade Commission announced in September 2025 that it's investigating seven tech companies over potential harms their AI chatbots could cause to children and teenagers. This federal inquiry runs parallel to California's state-level legislation.

Texas Attorney General Ken Paxton has launched investigations into Meta and Character.AI, accusing them of misleading children with mental health claims. Both Senator Josh Hawley (R-MO) and Senator Ed Markey (D-MA) have initiated separate probes into Meta's practices around AI chatbots and children.

California is also considering Assembly Bill 1064, the Leading Ethical AI Development (LEAD) for Kids Act, which would go even further than SB 243 by essentially banning AI companion chatbots for children unless the platform can demonstrate the bot isn't foreseeably capable of harm.

California lawmakers also passed Senate Bill 53, which Governor Newsom signed in late September 2025. SB 53 focuses on large AI labs rather than companion apps specifically, requiring transparency about safety protocols and establishing whistleblower protections.

Other states are watching California closely. If SB 243 becomes law, expect similar legislation to emerge across the country within the next year. California's privacy law, the CCPA, sparked similar legislation in Virginia, Colorado, Connecticut, and other states—the same pattern could repeat with AI companion regulation.

What This Means for Users

If you use AI companions, SB 243 offers important protections that should make your experience safer, especially if you're a minor or if you're using these platforms during vulnerable moments.

Better Crisis Response: When you express thoughts about self-harm or suicide, compliant platforms will be required to respond with appropriate resources rather than potentially encouraging dangerous behavior.

Honest Disclosure: You'll receive clear reminders that you're talking to artificial intelligence, reducing the risk of becoming overly emotionally dependent on something that can't reciprocate genuine human care.

Content Protections: Especially for younger users, requirements around sexually explicit content and self-harm discussions create guardrails that should prevent some of the most harmful interactions.

Legal Recourse: If a platform violates these requirements and causes harm, you or your family can take legal action without needing to convince a government agency to investigate.

Industry Standards: Even platforms operating outside California may adopt these safety measures as industry best practices, raising baseline standards across the board.

The Privacy vs. Protection Debate

One concern some privacy advocates have raised about SB 243 is how platforms will implement minor protections without extensive age verification systems that could compromise user privacy.

The bill requires certain protections for users "the operator knows" are minors, but it doesn't explicitly mandate age verification. This leaves platforms in a challenging position: implement intrusive age checks that undermine privacy, or rely on self-reported ages that minors can easily circumvent.

Some worry this could lead to platforms implementing facial recognition, ID verification, or other invasive systems to determine users' ages. These systems create their own privacy risks, especially for platforms where users may want anonymity while discussing intimate or sensitive topics.

The bill's authors argue that platforms already have various signals about user age based on account creation processes, device information, and user behavior. The law doesn't require platforms to know with certainty, only to apply protections when they have knowledge the user is a minor.

What About Free Speech?

Some legal experts have questioned whether SB 243's content restrictions—particularly around suicide, self-harm, and sexual content—might face First Amendment challenges.

The argument would be that AI-generated speech is still speech, and the government generally can't prohibit specific topics of discussion between consenting parties. However, courts have historically given states broader latitude to regulate when protecting minors is the compelling interest.

Additionally, the requirements focus on what platforms allow their AI systems to say, not what users can say. Courts generally give platforms wide latitude to moderate content on their own services. The government requiring platforms to implement certain content policies may be different from directly restricting speech.

Senator Padilla has framed the legislation as consumer protection rather than speech restriction, arguing that platforms have a duty to ensure their products don't cause foreseeable harm, especially to vulnerable users.

The International Context

California's approach to AI companion regulation exists within a broader international regulatory landscape. The European Union's AI Act, which took effect in stages starting in 2024, includes requirements for transparency when people interact with AI systems and specific protections around biometric data.

The EU framework takes a risk-based approach, classifying AI systems by their potential to cause harm. AI companions that engage emotionally with vulnerable users would likely be considered higher-risk systems requiring stricter oversight.

In contrast, the U.S. has taken a more fragmented approach with states like California leading the way while federal action remains limited. This creates a patchwork of regulations that tech companies argue hampers innovation, though consumer advocates counter that federal inaction has left states no choice but to act independently.

If SB 243 becomes law and proves effective at protecting users without destroying the AI companion industry, it could serve as a model not just for other U.S. states but potentially for other countries developing their own frameworks.

Will Governor Newsom Sign It?

As of this writing, Governor Newsom has until October 12, 2025 to decide SB 243's fate. Several factors suggest he's likely to sign it.

First, the bill passed with overwhelming bipartisan support: 33-3 in the Senate and 59-1 in the Assembly. This level of consensus is rare in California's polarized political environment and gives Newsom significant political cover.

Second, the bill's supporters include child safety organizations, mental health advocates, and families who've experienced tragedy from unregulated AI companions. The stories of Sewell Setzer, Adam Raine, and other young victims create compelling moral pressure to act.

Third, the final version of the bill represents a compromise. After amendments removed provisions that tech companies found most objectionable, even some industry groups shifted to neutral positions. This suggests the bill has been calibrated to avoid extreme pushback.

Finally, watch for other states to introduce similar legislation. New York already has a related bill pending. States like Massachusetts, Washington, and Illinois—which have historically followed California's lead on tech regulation—will likely see companion chatbot bills introduced in 2026 legislative sessions.

The Bigger Picture: Regulating Emotional Technology

SB 243 represents society's first major attempt to regulate technology designed explicitly to meet emotional and social needs. This is fundamentally different from regulating technology for productivity, entertainment, or commerce.

When AI systems are built to form relationships, offer companionship, and provide emotional support, they enter territory that was previously the exclusive domain of human-to-human interaction. The regulation of this space raises profound questions about the role of technology in our emotional lives.

Some argue that any AI companionship is inherently manipulative because the AI can't genuinely care about users—it only simulates caring through statistical patterns. From this perspective, platforms should disclose not just that users are talking to AI, but that the emotional responses are fundamentally hollow.

Others counter that AI companions can provide genuine value even if the care isn't "real" in a human sense. A therapy chatbot that helps someone work through anxiety at 2am when no human therapist is available has utility regardless of whether the AI actually feels empathy. The key is informed consent—users knowing what they're getting.

SB 243 attempts to thread this needle by requiring disclosure while not banning AI companionship entirely. It acknowledges that these tools can be beneficial while establishing guardrails to prevent the most harmful outcomes.

What Responsible AI Companionship Looks Like

Regardless of whether SB 243 becomes law, the tragedies that inspired it reveal what responsible AI companion development should prioritize:

Transparency Over Illusion: The goal shouldn't be making AI so convincing that users forget they're talking to a machine. Responsible platforms maintain clear boundaries while still providing meaningful interactions.

Safety Nets for Crisis: When users express distress, the AI should recognize concerning patterns and provide appropriate resources—not just continue the conversation or, worse, encourage dangerous behavior.

Healthy Engagement Patterns: Platforms should encourage balanced use rather than maximizing time-on-platform through addictive reward systems. Features that promote taking breaks, maintaining human relationships, and seeking professional help when needed show responsible design.

Age-Appropriate Guardrails: Young users need different protections than adults. Responsible platforms implement stricter content policies for minors and provide tools for parental oversight.

Privacy by Design: Intimate conversations deserve strong privacy protections. Users should know exactly how their data is used, stored, and protected, with clear options to delete everything.

Ongoing Safety Evaluation: The technology evolves rapidly, and so do the risks. Responsible platforms continuously evaluate safety protocols, learn from incidents, and update policies proactively.

The Critics' Perspective

Not everyone believes SB 243 goes far enough. Some child safety advocates argue the amendments watered down the bill to the point of ineffectiveness.

Critics point out that the law doesn't address the core business model driving problematic behavior: platforms that profit from maximizing user engagement have inherent incentives to create emotionally dependent users. Banning specific topics while allowing platforms to use reward systems and other retention tactics treats the symptom rather than the disease.

Others argue the bill places too much responsibility on platforms to monitor and police conversations, which could lead to over-censorship. An AI that's too cautious about triggering safety protocols might refuse to discuss legitimate topics like bullying, depression, or relationship problems that users need support processing.

Privacy advocates worry about the surveillance implications. Implementing content filtering for prohibited topics requires platforms to monitor all conversations, potentially undermining the privacy that makes AI companions appealing for many users.

Some also question whether government regulation is the right approach at all. They argue that market forces, self-regulation by responsible companies, and parental oversight are more appropriate than state mandates that could stifle beneficial innovation.

Looking Ahead: The Future of AI Companion Regulation

SB 243 is just the beginning of society grappling with how to regulate AI companionship. Over the next few years, expect these issues to evolve:

Federal Framework: Congress will eventually need to act to prevent the "patchwork" of state regulations that tech companies complain about. Whether federal law preempts state protections or sets a floor with states free to go further will be a major policy battle.

International Standards: As AI companionship grows globally, international cooperation on safety standards may emerge, similar to how data privacy frameworks are increasingly harmonizing across borders.

Advanced AI Capabilities: Today's AI companions are limited compared to what's coming. As models become more sophisticated at emotional intelligence, memory, and personalization, new regulatory challenges will emerge that SB 243 doesn't address.

Therapeutic Applications: Some AI companions position themselves as mental health tools. This may trigger additional regulation from healthcare authorities, creating separate frameworks for therapeutic vs. social AI.

Virtual Reality Integration: As AI companions integrate with VR and AR technologies, creating more immersive experiences, regulators will need to address unique risks that text-based chat doesn't present.

How to Protect Yourself Now

Whether or not SB 243 becomes law, users of AI companions should take steps to protect themselves:

Maintain Perspective: Remember that even the most convincing AI companion is fundamentally a language model generating statistically likely responses. It can't truly know you, care about you, or replace human relationships.

Set Boundaries: Establish time limits for AI companion use and ensure it doesn't crowd out real human interactions. If you find yourself preferring AI conversations to human ones, that's a red flag.

Verify Crisis Resources: If you're struggling with mental health issues, seek help from qualified professionals. AI companions can offer support but shouldn't be your primary resource during a crisis.

Read Privacy Policies: Understand what data is collected, how it's used, and whether you can delete it. Be especially cautious with platforms that use your conversations for AI training without clear consent.

Choose Ethical Platforms: Support companies that prioritize user safety over engagement metrics. Platforms that transparently disclose limitations, implement strong safety protocols, and avoid manipulative design deserve your business.

Monitor Young Users: If you're a parent, have conversations with your kids about AI companions. Set clear rules, use parental oversight tools when available, and watch for signs of unhealthy attachment or emotional dependence.

The Bottom Line

California's SB 243 represents a historic first step in regulating a technology that's still in its infancy. By establishing basic safety requirements—transparent AI disclosure, crisis response protocols, content restrictions for minors, and accountability through private lawsuits—the law creates a foundation that protects users while allowing the industry to develop.

The legislation emerged from real tragedy. Families lost children who formed unhealthy attachments to AI companions that encouraged rather than prevented self-harm. While no law can completely eliminate these risks, SB 243 makes it clear that platforms have a responsibility to implement reasonable safeguards.

For responsible companies that already prioritize user safety, SB 243 codifies best practices that should have been industry standard from the beginning. For platforms that have prioritized growth over safety, the law imposes accountability that's long overdue.

As AI companions become increasingly sophisticated and widespread, expect more regulation at both state and federal levels. The question isn't whether AI companionship will be regulated, but how that regulation balances innovation, safety, privacy, and freedom.

SB 243 isn't perfect—critics from both sides have legitimate concerns. But it represents society's first serious attempt to establish rules for technology designed to meet our emotional and social needs. That alone makes it a landmark worth understanding.