Senate Bill 867, signed by Governor Newsom, marks California’s first attempt to regulate AI-companion functionality built directly into children’s toys.
The law prohibits any person from manufacturing, selling, exchanging, or offering for sale any toy that includes a companion chatbot, with no available compliance path through disclosure or certification.
Unlike typical consumer protection legislation, SB 867 does not allow manufacturers to sidestep the ban by adding content filters, parental controls, or session time limits.
Avoiding the prohibition entirely requires companies to alter the product so its functionality no longer falls within the bill’s statutory definition of a covered companion chatbot.
That approach places the functionality itself at the centre of regulatory scrutiny, rather than treating safety disclosures or testing regimes as sufficient responses to the underlying risk.
The bill reflects a broader legislative trend, with New York’s Senate Bill 9051 seeking to prohibit chatbot operators from offering minors certain features deemed unsafe, and Senate Bill 9408 proposing a five-year moratorium on AI companion toys entirely.
California’s law also highlights the increasingly complex supply chains behind AI-enabled consumer products, where one company manufactures the device, another supplies the AI model, and an integrator connects the two.
Brand owners may control a chatbot’s persona and memory settings, while hosting providers deploy later software updates, and retailers ultimately place the finished product into consumers’ hands.
Failing to coordinate effectively at every stage of this chain could expose manufacturers, AI vendors, distributors, and retailers to significant litigation risk.
SB 867 does not declare that all software constitutes a product, and a statutory violation would not automatically establish a design defect under California law.
However, the law signals a meaningful shift in how regulators may frame AI-enabled consumer products, treating software-driven behaviour as part of a physical product’s safety profile rather than a separate digital service.
That framing could bring software behaviour within familiar product liability questions about who controlled the design, which risks were foreseeable, and whether a safer design was feasible.
Companies across a wide range of product categories, not just toy manufacturers, will want to proactively prepare for this regulatory shift as similar laws emerge across the United States.