California SB 243, often called the AI Companion Bill of Rights, is the most consequential piece of legislation focused on AI companion apps to be considered in any US jurisdiction. It was introduced by State Senator Steve Padilla in 2025 and has been through multiple iterations. The bill matters not just for what California requires of operators in California (a meaningful share of the market) but as a template other states will likely use.

This piece is a plain-language summary of what the bill requires, who is affected, where the legislative process stands, and what it means in practice.

If you have time for one paragraph: the bill sets out requirements for AI companion platforms operating in California, focused on disclosure (users must know they are talking to AI), mental-health resource integration (apps must provide crisis-line referrals when conversations indicate concerning patterns), age-related protections (additional requirements for minor users), and operator transparency. The specific provisions have evolved through committee markup; the core direction has been consistent. Whatever passes will become the floor for industry practice across the US, because most companies will not maintain different products for California versus elsewhere.

What the bill is responding to

SB 243 was developed in part in response to the Garcia v. Character Technologies wrongful-death lawsuit and to broader concerns about AI companion app effects on minors and on users in mental-health crisis. The legislative record shows committee testimony from researchers, advocacy groups, and the Setzer family. We covered the lawsuit in a separate piece.

The broader regulatory environment includes related work at the federal level (FTC inquiries, FCC rule-making in adjacent areas), at the EU level (the AI Act applies to companion apps among other AI systems), and in other US states (similar bills have been introduced in New York, Massachusetts, and elsewhere, often modeled on California’s approach). California is the leader; the others are watching.

What the bill requires

The provisions have evolved as the bill moved through committee. As of the most recent public version, the bill includes requirements in five general areas. Specific text and exact provisions should be verified against the current bill text at leginfo.ca.gov.

Disclosure that the companion is AI. Apps must clearly disclose to users that they are interacting with an AI rather than a human. The requirement extends across the user experience, not just at signup. The aim is to prevent a user from forgetting or being confused about the nature of the system they are talking to.

Mental-health resource integration. Apps must provide crisis-line referrals (988 in the US) and other appropriate mental-health resources when conversations show patterns indicating mental-health risk. The specifics of what triggers a referral and how it is presented are subject to implementation rules that would follow the bill’s passage.

Protections for minor users. Apps must implement reasonable age verification, restrict certain content for users determined to be minors, and meet additional disclosure and intervention requirements when a minor is identified as the user. The specifics on age verification (which is technically and politically contested) are part of the open implementation question.

Operator transparency. Apps must publish certain information about how their systems work, what data is collected, and how it is used. The specifics are similar in spirit to the disclosure provisions of the EU AI Act, with California-specific tailoring.

User rights and remedies. Users have rights to information about their data, to deletion of their data in certain circumstances, and to remedies if the operator fails to comply. The civil penalties and enforcement mechanisms are part of the bill’s specifics.

Where the bill stands

Bills move; this section will date quickly. Verify against the current legislative status at the California legislature’s website (leginfo.ca.gov, where you can search SB 243 directly).

As of this writing, the bill has been through committee markup, has had public testimony from multiple stakeholders, and is in active consideration. Whether and in what form it passes, when it would take effect if passed, and how the implementation rules would be developed are all open. Industry response has been substantial; advocacy-group response has been substantial; the legislative arithmetic has been close.

If the bill passes, an implementation period would follow, during which operators would adjust their products to comply. The exact timeline depends on the bill text as enacted.

What this means for operators

If you operate an AI companion app, the practical near-term implication is that the design principles emerging from SB 243 are likely to become industry table stakes regardless of whether the bill passes in its current form. Most major operators are already moving in this direction. Specifically:

  • Visible AI-disclosure throughout the user experience, not just at signup
  • Crisis-resource integration that triggers on patterns rather than only on explicit requests
  • Some form of age verification or age-appropriate-design choices for minor users
  • Public-facing privacy and data documentation

The defensible product position is to build for the eventual regulatory environment rather than the current minimum. Several of the apps we cover (Kindroid, Pi) are already doing most of this; the gap to compliance for them would be small. Other apps will face more significant changes.

What this means for users

Practical implications, mostly modest in the short term and more significant in the longer term.

Crisis-resource visibility will increase. You will see crisis lines and mental-health resources surfaced more prominently in companion apps over the next year or two. Some of this will feel like product friction; the framing is that this is part of how the industry handles the responsibility of being a place users sometimes go in difficult moments.

Age-related friction will increase, especially for adult-content apps. Age verification is the place this regulation will be most visible. Expect more apps to require additional steps at signup, particularly for adult-capable apps.

Disclosure will become more visible. Apps already nominally disclose that they are AI; the regulation pushes this to be more prominent and more frequent. Users who have built relationships with their AI companions will see the disclosure language more often than they currently do.

The product environment will become more conservative around mental-health-adjacent topics. Apps will be more cautious about engaging with topics where misjudgment could lead to harm. For most users in most situations, this will be invisible. For users specifically using AI companions in mental-health-adjacent contexts, the apps may feel less helpful for those use cases.

What this does not change

The bill regulates platform behavior, not user behavior. The deeper questions about what AI companion use does to users over the long term are not what regulation can answer; they are what research has to address. We covered the research backdrop here.

The bill also does not address several things people sometimes hope it would: it is not a content-moderation framework for adult material on AI platforms, it is not a copyright regime for AI training data, it is not an attempt to restrict the existence of AI companion apps. The scope is specifically the disclosure, mental-health, age, and transparency requirements above.

The Garcia v. Character Technologies lawsuit is the parallel legal track to SB 243’s regulatory track. The two are related but distinct. We covered the lawsuit in its own piece.

The EU AI Act classifies AI companion apps as one category of AI system subject to general transparency and disclosure requirements. The specifics of how the AI Act applies to companion apps are still being clarified through implementation guidance.

The Italian Garante’s actions on Replika in 2023 are an earlier example of European data-protection authorities engaging with this category. We covered this in Replika and the Italian Garante.

FTC inquiries in the US have looked at consumer-protection aspects of AI companion apps, including marketing claims and data practices. No specific rule-making has yet emerged but several investigations have been reported.

Other state bills modeled on California’s approach have been introduced in New York, Massachusetts, and elsewhere. Whether and how these progress depends partly on what California does.

FAQ

Has SB 243 passed?

Verify the current legislative status at leginfo.ca.gov. The bill has been in active consideration; final passage and effective date depend on the legislative process.

Does this regulate apps that are not based in California?

If they have users in California, generally yes. The question of jurisdiction over out-of-state and out-of-country operators is its own legal complexity, but most major operators will have to comply for their California user base, and most will not maintain different products for California versus elsewhere.

Does this affect adult companion apps differently?

Probably yes, especially on age verification. The general provisions apply across the category; the age-related provisions are most consequential for adult-capable apps.

Will this kill the AI companion industry?

No. The regulation is in line with what most of the industry is already moving toward. The smaller and less-resourced operators may struggle with compliance; the major players have largely been preparing.

Is this the EU AI Act?

No, separate. The EU AI Act covers AI systems generally and has its own provisions for companion-style systems. SB 243 is California-specific and AI-companion-focused. They share spirit on disclosure and transparency.

Where to read the primary documents

The bill text is at leginfo.ca.gov, searchable as SB 243. The legislative record (committee analyses, public testimony, vote records) is also available there. The most useful reading is often the committee analyses, which summarize the bill’s effects and the arguments for and against.

Garcia v. Character Technologies for the parallel legal track.

Replika and the Italian Garante for the earlier European regulatory action.

AI Companions and Mental Health for the research backdrop.

Editorial standards for our policies on related topics.

If you are tracking the bill or the broader regulatory landscape and have something we should know, write us at the contact form.