California has just wrapped up its 2026 legislative session with another wave of laws mandating how children interact with social media and artificial intelligence (AI) companion chatbots. Compared with previous years, the state took a more prescriptive approach, heavily regulating online service design before enough information was available to determine whether the lighter rules passed in recent years work.
California wields tremendous influence in setting the nation’s tech policy agenda, with other states often passing bills that mirror California’s. The 2026 session provides a few positive examples to follow, including protections for students and for adults seeking to remove childhood videos their parents posted and monetized online. However, the broader trend toward heavier regulations and design mandates is concerning.
In 2024, California’s Senate Bill 976 let parents decide whether their children could use personalized social media feeds by providing consent. It also gave them tools to limit notifications and time spent on those feeds. This year’s Assembly Bill 1709 removes personalized feeds for children under 16 entirely, even when parents supervise and consent to their use, while preserving that option for 16- and 17-year-olds. The law also prohibits autoplay—the feature that automatically plays videos as users scroll—for younger users and expands the attorney general’s authority to extend restrictions to additional features and platforms in the future, overriding parents’ judgment about whether these features benefit their child and what limits are appropriate.
Companion chatbots
California is taking the same aggressive approach with AI companion chatbots. Less than a year ago, the state passed Senate Bill 243, which required chatbots to disclose they are AI, maintain self-harm safeguards, and remind kids to take breaks. But before those rules and their impacts could fully take hold, the state passed Adam’s Law (Senate Bill 1119), which considerably expands these requirements.
The new law requires companies to erase a chatbot’s conversational memory and cuts off a child’s access after one hour per session (two hours a day total), unless a parent actively changes the settings. Developers must also conduct risk assessments, follow crisis-response procedures, publish incident reports, and undergo independent audits. Some reporting and crisis requirements already existed under last year’s law, but the new obligations add compliance costs without necessarily improving safety. These requirements increase the resources companies must devote to documenting compliance, while their value depends on whether the findings lead to measurable improvements in chatbot safety.
The law also requires companies to take “reasonable steps” to prevent chatbots from offering excessive flattery or encouraging emotional dependence. But it is difficult to define what counts as excessive praise or when a supportive conversation becomes unhealthy, let alone enforce it consistently. Developers will have to anticipate how regulators interpret these terms and adjust their chatbots accordingly, even when the responses themselves do not cause any harm.
California went further with Senate Bill 867 by banning the sale and manufacture of children’s toys that contain companion chatbots until 2031. By the time the ban expires, the technology will look entirely different. Even manufacturers that develop stronger safety features will be unable to sell their products in California, potentially jeopardizing their viability for the U.S. market as a whole. Meanwhile, companies operating in markets where these toys remain legal can continue improving them, testing safeguards, and learning from how families actually use them. Other markets will spend those four years learning how to make these toys safer and better, while California has legally locked its own creators out of the workshop.
Other chatbot proposals failed to become law. Assembly Bill 1988, the PAUSE Act, would have forced companies to freeze a chatbot conversation the moment a user repeatedly expressed being in a crisis, locking the chat until a human could review it. Such intervention might help in some cases, but abruptly interrupting a distressed person’s conversation could also exacerbate a crisis or discourage them from seeking further support. Assembly Bill 2023 and Senate Bill 300, which proposed additional chatbot safety requirements, did not receive a full chamber vote.
Gov. Gavin Newsom also vetoed Senate Bill 903, which would have restricted AI use in psychotherapy. In his veto letter, he warned that forcing clinicians to directly approve routine screening determinations would drastically limit tools already helping deliver care.
Age assurance and privacy
Last year, California passed Assembly Bill 1043 to build a device-level alternative to invasive age verification. Rather than demanding government identification documents, the law let parents set an age bracket on a child’s phone. The law required operating system providers to create the process for parents to set this age. However, many raised concerns that this requirement was overly broad and would wrap up traditionally open-source operating systems like Linux. This year’s Assembly Bill 1856 narrows the scope of the operating system definition to exclude such systems and maintain user choice.
California also revisited its 2022 Age-Appropriate Design Code, which required online services likely to be used by children to adopt strong privacy settings and assess potential harms to young users. A federal court blocked key provisions, finding that the required assessments likely violated the First Amendment by compelling companies to evaluate potentially harmful content. This year’s Assembly Bill 2246 keeps many of the original privacy requirements, such as limiting children’s data collection and turning off location tracking by default. To address the court’s objections, the bill removes the mandatory assessments, revises vague language about children’s welfare, and expressly disclaims any duty to restrict lawful speech. But it also requires companies to take reasonable steps to prevent serious harm to children, including psychological harm, which is difficult to predict or link to a particular design feature. That uncertainty also raises the stakes, as the new law doubles penalties and removes the previous law’s 90-day opportunity to fix violations.
Social media liability
Under Assembly Bill 2, large social media platforms face much higher damages if their negligence causes injury to a child. California law already allowed families to sue platforms for negligence, so the bill doesn’t create a new duty of care. Instead, AB 2 lets plaintiffs seek $5,000 per violation, up to $1 million per child, or three times actual damages, whichever is greater. Families must still prove that a company’s negligence caused the injury, but the law leaves it unclear how to count individual violations. When damages can accumulate this way, platforms have a stronger incentive to limit potential liability, which could mean restricting younger users’ access rather than offering more meaningful and flexible controls.
Assembly Bill 302 prevents schools from requiring students to use personalized social media feeds to participate in extracurricular activities or making those feeds their only communication channel. For example, a school using a social media website to distribute team announcements or club updates would need to provide another communication channel, such as email or a school portal. This preserves parents’ ability to keep their children off those feeds without sacrificing participation in school activities or access to school information.
Senate Bill 1247 gives adults who appeared in monetized family videos as children a way to request their removal or editing. California already required compensation protections for child influencers—children featured in money-making social media content managed by their parents— but those children had no specific right to remove videos their parents had monetized once they became adults. This new law gives them greater control over the commercial use of their childhood images.
California’s 2026 legislative session shows how quickly the state is moving from parental consent and basic safety safeguards toward more restrictive rules governing how children can operate online. Many of these new requirements build on laws passed just a year or two ago, before their effectiveness has been tested. Some measures, particularly the state’s privacy-preserving age-signal system and new protections for students and former child influencers, offer useful models for other states.
But the broader shift toward mandatory restrictions and greater liability raises concerns about the state’s trend toward overruling parental choice about device use in the home. As other states consider following California’s lead, they should examine whether these additional requirements actually improve children’s safety, improve consumer welfare, or protect user privacy before adopting rules that limit access and innovation.
For a complete look at all of this year’s tech regulations, check out our tracker: Roundup of California’s 2026 Technology Bills.