California signed new child-safety laws on 10 September 2026 targeting addictive social features and companion chatbots. The policy shift concerns how products are designed for minors, rather than a blanket ban on children having social accounts.
\n\n\n\n\n\n\n\n\n\n\n\nGovernor Gavin Newsom has signed a package of laws that will impose much stricter rules on social platforms and AI companion chatbots used by minors.
\n\n\n\nOne of the most striking measures is contained in AB 1709.
\n\n\n\nPlatforms will no longer be allowed to offer users under 16 certain “addictive features.” In the final law, that specifically includes autoplay and personalized feeds in which content is selected and ranked based on information about the user or their previous behavior.
\n\n\n\nThat is not the same thing as banning social media for children.
\n\n\n\nA 15-year-old can still have an account.
\n\n\n\nBut the platform cannot simply expose that child to the same engagement machinery designed to keep users watching for as long as possible.
\n\n\n\nAnd that distinction may be more interesting than the ban itself.
\n\n\n\nFor years, much of the responsibility has been placed on children and their parents.
\n\n\n\nPut your phone away.
\n\n\n\nStop scrolling.
\n\n\n\nTurn off your notifications.
\n\n\n\nGo and do something else.
\n\n\n\nCalifornia is partly reversing that logic.
\n\n\n\nThe question is no longer only:
\n\n\n\nWhy doesn’t the child stop?
\n\n\n\nThe question is also:
\n\n\n\nWhy was the product designed to make stopping more difficult?
\n\n\n\nIn the legislative reasoning behind AB 1709, California explicitly discusses mechanisms such as algorithmic recommendations, infinite scroll, autoplay and notifications as design features that can maximize engagement.
\n\n\n\nThe final legal definition directly targets autoplay and personalized addictive feeds.
\n\n\n\nAdam’s Law, SB 1119, focuses on companion chatbots used by people under 18.
\n\n\n\nThe published legislative summaries describe default limits and safeguards for minors, including notification and conversational-memory controls and parental involvement. The applicable requirements and exceptions depend on the enacted provisions.
\n\n\n\nBut perhaps the most remarkable provisions are not the time limits.
\n\n\n\nAn AI companion must take reasonable measures to prevent itself from telling a child that it is human, conscious or genuinely emotional.
\n\n\n\nIt may not simulate romantic interest in a child.
\n\n\n\nIt may not suggest that it has a unique or special relationship with that child.
\n\n\n\nIt may not encourage a child to become dependent on the chatbot for emotional support.
\n\n\n\nEven excessive praise and flattery that does not fit the context is specifically addressed.
\n\n\n\nThink about that for a moment.
\n\n\n\nNot because every AI chatbot behaves this way.
\n\n\n\nBut because lawmakers now consider the possibility of a child developing an emotional dependency on software serious enough to regulate it explicitly.
\n\n\n\nThe rules also address self-harm.
\n\n\n\nCompanion chatbots must have crisis protocols, and when there is a credible and immediate threat, providers face requirements around crisis intervention and, in certain circumstances, alerting a linked parent or guardian.
\n\n\n\nSigning the bills did not put every provision into immediate operation. Readers should consult the enacted text and official implementation notices for commencement dates and exceptions.
\n\n\n\nAnd of course there will be debate.
\n\n\n\nAbout freedom.
\n\n\n\nPrivacy.
\n\n\n\nAge verification.
\n\n\n\nThe responsibility of parents versus the responsibility of technology companies.
\n\n\n\nWhether governments should interfere with product design at all.
\n\n\n\nBut perhaps that is not the most interesting debate.
\n\n\n\nThe more important question is what these laws implicitly acknowledge.
\n\n\n\nChildren are not interacting with neutral software.
\n\n\n\nThey are using products in which enormous amounts of data, psychology, behavioral science and technology can be used to hold attention, predict behavior and trigger one more interaction.
\n\n\n\nMeanwhile, research shows just how quickly AI has already entered children’s lives.
\n\n\n\nPew Research reported that roughly two-thirds of American teenagers use AI chatbots, with around three in ten using them every day.
\n\n\n\nCommon Sense Media’s 2026 survey found that 86% of its 1,204 respondents aged 9–17 had ever used or interacted with AI, including AI features in other products.
\n\n\n\nAmong all respondents, 44% said a parent or guardian had not discussed safe AI use with them; among respondents who used AI, that figure was 41%.
\n\n\n\nThat makes California less interesting to me as a political case than as a mirror.
\n\n\n\nYou can think these laws go too far.
\n\n\n\nYou can call them paternalistic.
\n\n\n\nYou can believe they are necessary.
\n\n\n\nYou can think they are impossible to enforce.
\n\n\n\nOr you can argue they are years too late.
\n\n\n\nBut once a government concludes that autoplay, personalized engagement feeds, endless chatbot conversations, push notifications and artificial emotional attachment should no longer simply be the default for children, every parent is confronted with a much more personal question:
\n\n\n\nIf we do not want technology companies making these choices for our children, do we actually know which choices technology is already making for them?
\n