California Governor Gavin Newsom signed a package of bills on Thursday that reaches into two of the state’s most visible technology markets, restricting addictive social media features for users under 16 and limiting how minors interact with AI chatbots.
The two bills aimed at children are the headline. One restricts addictive features on social media platforms for users younger than 16. The other limits interactions between minors and chatbots and establishes legal liability when a chatbot causes harm. Both drew public support from Anthropic and OpenAI, the two AI companies that have been most vocal about backing regulation.
The support is strategic. California is the most active state in the country on AI legislation, and OpenAI has backed a series of state bills, including SB 813 and AB 1405, while calling for national safety rules. Standing on the side of regulation is cheaper for large companies than for startups: compliance costs are fixed, and the biggest firms already have the staff and systems to meet them.
Newsom’s signature follows a more complicated history on AI. In September 2024 he vetoed SB 1047, the sweeping safety bill that would have required developers of the largest models to run safety tests and accept liability for catastrophic harm, arguing it targeted only the biggest systems while leaving the rest untouched. The veto drew relief from much of the industry and criticism from some safety advocates. What Newsom has signed instead is a series of narrower measures, a pattern the two new bills extend.
The chatbot liability bill is the more novel piece. Lawmakers in several states have discussed holding AI companies responsible when their products cause harm, but California’s bill turns that discussion into law, and it does so for the most sensitive population. The details of how liability will be proven will matter as much as the principle.
California is not the first state to regulate AI, but it is the one that matters most. Colorado enacted a broad state AI law in May 2024, Tennessee passed the ELVIS Act that same year to shield musicians’ voices from unauthorized cloning, and other states have taken narrower slices. What sets California apart is scale: rules written in Sacramento tend to become the default for companies that cannot afford a different product for every state.
The social media bill follows years of state and federal effort to curb features that keep young users scrolling. California has already enacted rules on age-appropriate design, and this package extends that framework to the addictive mechanics themselves rather than just the data practices behind them.
The addictive-design fight has its own California lineage. The state passed an Age-Appropriate Design Code in 2022 requiring platforms to weigh children’s well-being, and a federal court blocked it the following year after NetChoice, the industry trade group, challenged it. Congress has not finished the job either: the Kids Online Safety Act cleared the Senate in 2024 but stalled. The new bill tries again at the level the courts left open, targeting the features themselves.
Analysts said the laws will face the same practical question that has slowed similar efforts: enforcement. Age verification on the internet is unsettled, and platforms have repeatedly challenged such rules in court. The new bills give the state another tool, but their reach depends on how they survive legal review.
The industry’s own posture has shifted, which is part of why the bills drew company support. More than forty state attorneys general sued Meta in 2023 over features they said were designed to hook young users, and the US surgeon general issued an advisory the same year linking heavy social media use to worse mental-health outcomes in adolescents. Facing that pressure, the largest platforms have started asking for rules they can plan around rather than fighting each one after the fact.
The bills also reflect a bet that state action will come ahead of federal rules. Congress has debated AI safety for years without passing a statute, and the companies themselves have asked for a single national standard, if only to replace a patchwork of state laws. Until that happens, Sacramento’s version becomes the default for the country’s largest consumer-technology market, and other states tend to copy what California enacts rather than write their own from scratch.
For the AI companies, the calculation is explicit. Anthropic and OpenAI have concluded that a patchwork of state rules is preferable to the vacuum, and that helping write the rules beats having them written by opponents. Their endorsements give the bills bipartisan cover and a credibility that the industry’s earlier, more defensive posture lacked.
The signing also positions California as the de facto national regulator on AI, a role the state has claimed by moving faster than Congress. Companies operating nationwide often build to the California standard, which means laws signed in Sacramento shape products used everywhere.
The two AI firms and the governor now share a stake in how the rules are enforced. The laws are signed; the test is whether they change what a sixteen-year-old sees on a screen, and what a chatbot does when that user asks it for help.
