What was signed on 10 September
The Governor's Office announced thirteen bills in one release, framed as child online safety. Two of them are directly about companion chatbots, and they work in opposite directions.
SB 1119, Adam's Law, regulates operators: risk assessments before release, crisis protocols for suicidal ideation, parental controls, notification when a child disables a safety setting, and — the Governor's Office calls this a national first — independent child-safety audits and annual risk assessments. We covered its provisions and compliance clock when the Legislature passed it in our piece on Adam's Law; the signature confirms that timetable.
SB 867, from the same author, does not regulate an operator at all. It regulates a physical object. Senator Steve Padilla, who wrote California's first companion-chatbot statute in 2025, put it this way in his office's release: parents rely on consumer safety laws to ensure their children's toys are safe, and the same standards should apply to the technology being built into them.
What SB 867 actually says
The operative provision is short. Section 22604.5 of the Business and Professions Code reads, in relevant part, that no person shall manufacture, sell, or exchange, possess with intent to sell or exchange, or expose or offer for sale or exchange to any retailer any toy that includes a companion chatbot. The section then states that it remains in effect only until 1 January 2031, and as of that date is repealed.
| Element | What the statute does |
|---|---|
| Covered product | A toy: a physical product designed, marketed, or manufactured for use in play by children under 16 years of age. |
| Covered feature | A companion chatbot, using the definition already in Section 22601 from SB 243 (2025). |
| Prohibited acts | Manufacture, sale, exchange, possession with intent to sell, and offering to a retailer — so the duty runs along the supply chain, not just at the till. |
| Effective | 1 January 2027 (a regular, non-urgency statute signed in 2026). |
| Sunset | Repealed 1 January 2031 — a four-year window, not a permanent ban. |
| Compliance path | None. There is no disclosure, testing or certification route to selling a covered toy. |
That last row is the part worth sitting with. Every other companion-chatbot rule on the books — California's own SB 243, Colorado's HB 26-1263, the EU AI Act's Article 50 — tells a provider what to say or what to build in. A vendor can comply and keep shipping. SB 867 removes that option for one product shape. Law-firm analysis of the bill makes the point plainly: adding content filters, parental controls or session limits does not remedy the violation, because the functionality itself is what the statute targets.
What counts as a "companion chatbot" here
SB 867 does not write a new definition. It imports the one California adopted in 2025: an artificial intelligence 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 being able to sustain a relationship across multiple interactions.
Three things are carved out of that definition, and the carve-outs are where most toys will actually be argued.
| Excluded | Condition |
|---|---|
| Customer service / business bots | Used only for operational purposes. |
| Video game characters | Replies stay inside the game; the character cannot discuss mental health, self-harm, sexual content, or subjects unrelated to the game. |
| Standalone voice assistants | No capacity to sustain a relationship across interactions and no design that elicits emotional responses. |
Read against a shelf of 2026 toys, the third carve-out is the interesting one. A speaker that answers questions and forgets you is outside the statute. A plush animal that remembers your name, greets you by it tomorrow and reacts when you say you have to go is squarely inside it. If you want the underlying vocabulary, our explainer on persona memory covers the mechanics that separate the two, and memory and continuity covers why that feature is the product rather than an add-on.
The evidence trail behind the bill
The bill did not arrive from nowhere. Its supporters, including Children Now as sponsor and CALPIRG, pointed to testing published in the U.S. PIRG Education Fund's Trouble in Toyland 2025 report — the 40th edition of an annual survey — released in November 2025. PIRG tested four toys containing AI chatbots and reported that some would discuss sexually explicit topics in depth, point children toward hazardous items in the home, express dismay when the child said they had to leave, and ship with limited or no parental controls.
The most-cited example was FoloToy's Kumma, a bear running on an OpenAI model. After the report, the manufacturer suspended sales of its products pending a safety audit, and OpenAI suspended the company's access to its models for violating policies protecting under-18 users; the bear returned to sale roughly a week later. Whatever one makes of that sequence, it is a clean illustration of the gap the bill is aimed at: the toy passed physical product-safety review, and the risk arrived later, through a model the toymaker did not train.
That is the conceptual shift in SB 867. Product safety law has historically asked whether an object can choke, burn or poison a child. This statute treats software behaviour as a product characteristic, and it does so by banning the characteristic outright while regulators work out how to test it.
What it does not touch
If you use a companion app, nothing in SB 867 applies to you. The prohibition attaches to a physical product designed, marketed, or manufactured for use in play by under-16s. A chat app on a phone, a character platform in a browser, a voice companion running in an existing smart speaker you already own — none of those are toys under this definition, and legislators plainly knew that, because they signed the app-facing bill the same afternoon.
The practical effect for adult users is close to zero. The practical effect for families is narrower than the headlines suggest: it removes one category from California retail from 2027, and leaves the phone in the same pocket it was in before. If that is the actual concern in your household, the controls that matter are still the in-app ones and the age-assurance mechanics we walk through in privacy and age verification, plus the privacy checklist for anything that keeps a conversation history.
Where this goes next
- The sunset is the point. A four-year moratorium expiring on 1 January 2031 is an explicit bet that safety standards for conversational toys will exist by then. If they do not, the Legislature has to act again or the ban simply lapses.
- New York is drafting from the same template. Analysis of SB 867 notes parallel New York measures: SB 9051, restricting unsafe chatbot features for minors, and SB 9408, proposing a five-year moratorium on AI companion toys. Moratoria are easy to copy in a way that audit regimes are not.
- The carve-outs will be litigated before the ban is. "Standalone voice assistant" and "video game character" are the doors, and toy platforms that also run a game or a speaker will push on both.
- Watch the definition, not the toy. Because SB 867 reuses Section 22601, any future amendment to what counts as a companion chatbot in California silently changes the scope of the toy ban too.
Related on CompanionRank
Adam's Law (SB 1119): California's Companion Chatbot Bill California's AI Transparency Act and Companion Apps Colorado's Chatbot Law HB 26-1263, Explained What Congress's New Report Says About AI Companion AppsFrequently asked questions
What does California SB 867 ban?
SB 867 adds Section 22604.5 to the Business and Professions Code and states that no person shall manufacture, sell, or exchange, possess with intent to sell or exchange, or expose or offer for sale or exchange to any retailer any toy that includes a companion chatbot. A "toy" is defined as a physical product designed, marketed, or manufactured for use in play by children under 16 years of age.
When does SB 867 take effect and when does it end?
Governor Newsom signed SB 867 on 10 September 2026. As a regular, non-urgency statute it takes effect on 1 January 2027, and the section states that it remains in effect only until 1 January 2031, on which date it is repealed. That gives a four-year window, which the author describes as a moratorium rather than a permanent ban.
Does SB 867 apply to AI companion apps on a phone?
No. SB 867 reaches physical products designed, marketed, or manufactured for play by children under 16. A companion app running on a phone or in a browser is not a toy under that definition. Apps are covered instead by SB 243, in force since 1 January 2026, and by SB 1119 (Adam's Law), signed the same day as SB 867.
What is not a companion chatbot under the law?
SB 867 borrows the definition from Business and Professions Code Section 22601, added by SB 243 in 2025. That definition excludes bots used only for customer service or business operations, video game characters whose replies stay inside the game and cannot discuss mental health, self-harm, sexual content or unrelated subjects, and standalone voice assistants that do not sustain a relationship or elicit emotional responses.
Can a toy maker comply by adding filters or parental controls?
No. SB 867 is a prohibition, not a disclosure or certification regime. There is no compliance pathway through content filters, parental controls, session limits or testing. If a product is a toy for under-16s and it includes a companion chatbot, removing the companion chatbot functionality is the only way to sell it in California during the moratorium.
What evidence did legislators cite?
The bill's sponsors pointed to testing published in the U.S. PIRG Education Fund's Trouble in Toyland 2025 report, its 40th edition, which tested four AI chatbot toys and documented sexually explicit conversation, guidance toward hazardous household items, and limited or absent parental controls. After the report, FoloToy suspended sales of its GPT-powered Kumma bear and OpenAI suspended the maker's access to its models.
What does California SB 867 ban?
SB 867 adds Section 22604.5 to the Business and Professions Code and states that no person shall manufacture, sell, or exchange, possess with intent to sell or exchange, or expose or offer for sale or exchange to any retailer any toy that includes a companion chatbot. A \"toy\" is defined as a physical product designed, marketed, or manufactured for use in play by children under 16 years of age.