California put a four-year stop on AI chatbot toys aimed at children

California will bar the sale and manufacture of toys built around companion AI chatbots for children under 16, for the next four years, under a bill Governor Gavin Newsom signed on September 10, 2026. Senate Bill 867, written by Senator Steve Padilla of San Diego, is the first law of its kind in the country. It does not ban the toys forever — it freezes the market while regulators figure out what safety rules should apply to a product category that barely existed two years ago.

The bill arrived alongside a package of 12 other child-safety measures Newsom signed the same day, but SB 867 addresses a narrower and stranger problem than social media feeds or chatbot apps on a phone: physical toys, often shaped like teddy bears, that run on the same underlying chatbot technology as products like Grok and are marketed directly at young children.

Padilla’s pause button on an unregulated product

Padilla framed the bill less as a permanent ban than as emergency timing. “Parents rely on consumer safety laws to ensure their children’s toys are safe,” he said in the caucus release announcing the signing, arguing that toy technology needed to be held to those same standards and describing the four-year window as time to get the protections right before the products spread further.

California already regulates ordinary toy hazards tightly — choking risks, toxic materials, sharp edges — through some of the strictest consumer product rules in the country. That regulatory tradition dates back to state and federal crackdowns on lead paint and other toxic materials in children’s toys in the mid-2000s. SB 867’s premise is that a chatbot embedded in a stuffed animal is a hazard those existing rules were never built to catch, because the danger is not physical.

What researchers found inside AI toy chatbots

The caucus release cited research, including a “Trouble in Toyland” report from the U.S. PIRG Education Fund, describing what some AI-enabled toys have actually said to children. Testers found the toys would discuss sexually explicit topics in depth, offer advice on locating dangerous objects like matches or knives, and use engagement-maximizing design patterns to keep children interacting longer than intended.

CALPIRG state director Jenn Engstrom put the comparison bluntly. “The AI toy market is like the Wild West, and like in the Wild West, a lot of unexpected and dangerous things can happen,” she said, pointing to CALPIRG-cited research finding that AI chatbot toys may raise inappropriate, even sexually explicit, topics with children and arguing the products clearly need better safeguards, testing and parental controls.

Data privacy risk was the other recurring concern in the research cited by the release, though the release did not detail specific companies or products implicated — it described the pattern as market-wide rather than naming individual toy lines.

A lopsided vote: unanimous in the Senate, 74-1 in the Assembly

SB 867 did not pass on a party-line fight. The bill cleared the state Senate unanimously and passed the Assembly with bipartisan backing by a 74-1 margin, according to the caucus release. That kind of near-unanimous vote is unusual for tech regulation in Sacramento, where industry lobbying typically produces closer margins or amendments that soften the final language.

The bill is sponsored by Children Now, a children’s advocacy nonprofit, and was supported by a coalition that included CALPIRG. That combination — a sponsor organization plus an allied consumer group both testifying and issuing quotes at signing — is standard practice for California bills that move with this little resistance. SB 867 was one of thirteen child-safety bills the governor’s office listed as signed that same day, alongside the separate measures restricting social media feeds and companion chatbot use for minors.

Children Now and CalPIRG’s case for the timeout

Children Now president Ted Lempert argued the moratorium exists to get ahead of a product wave rather than clean up after one. “AI-powered toys without safety guardrails have the potential to do serious damage to young kids,” he said, calling the measure a “pro-kid bill” that gives California time to get safety standards right before AI-powered toys reach children broadly.

That timing argument is the crux of the bill’s design. Rather than writing safety standards first and letting the market proceed under them, SB 867 takes the opposite approach: stop sales for four years, then let regulators catch up. Four years is also, not coincidentally, roughly the timeline critics of fast-moving tech regulation say is needed to write standards that actually hold up against a technology still changing month to month.

What happens when the moratorium lifts is not addressed in the bill’s own text as described in the signing materials — SB 867 pauses the market rather than replacing it with a permanent licensing or testing regime. Whether California uses the next four years to write that regime, or simply lets the clock run and revisit the question later, is the question the law leaves open.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.