Thirteen new California laws now place some of the nation’s most extensive child-safety duties on social-media and artificial-intelligence companies, shifting a growing share of responsibility for young people’s digital environments away from families and schools and toward the businesses that design them. Gov. Gavin Newsom signed the package on September 10, according to an AP account of the measures, which restrict engagement-driven feeds for children under 16, impose new controls on companion chatbots, strengthen liability for harm, revise age-assurance rules and give families greater authority over school-issued devices.

For educators, the package is consequential even though most of its requirements fall on technology companies rather than school districts. Schools have spent years trying to govern student behavior at the point of use: limiting phones, filtering networks, writing acceptable-use rules and teaching media literacy. California is now testing a different premise—that the architecture of a product can create risks that cannot be managed solely through better choices by children, parents or teachers.

The laws do not settle scientific or constitutional disputes surrounding youth technology. They turn those disputes into implementation questions about harmful features, age determination, audit evidence and access to beneficial online communities. Because California is both the most populous state and the home of many technology companies, the answers may influence products and school practices far beyond its borders.

A Rule for Design, Not Simply Screen Time

Assembly Bill 1709 targets specific mechanisms used to sustain attention rather than setting a daily time limit. Its statutory language defines addictive features to include algorithmic feeds and autoplay, then bars covered platforms from providing them to users younger than 16. A platform may still allow a young person to maintain an account if those features are removed, a distinction that makes the law less a categorical prohibition than a mandate to redesign the experience.

That distinction separates access to information from the mechanics that continuously recommend it. A chronological feed, a requested video or a private message can present different risks from an endless stream personalized around prior engagement. The law therefore treats interface design as a variable that can be regulated, not as a neutral container for content.

Compliance will depend on age assurance. Platforms must verify a user’s age before supplying restricted features, while related legislation updates the state’s system for transmitting age-bracket signals from operating systems or application stores. The enrolled AB 1856 summary says those signals are to be requested only when required by law, an attempt to prevent age information from becoming a general-purpose tracking tool.

The enforcement provisions are substantial: AB 1709 authorizes civil penalties of as much as $50,000 per affected minor for a knowing violation and $25,000 for a negligent one. It also creates an advisory commission with expertise spanning pediatrics, psychology, technology, civil society, academia and K–12 education. That multidisciplinary structure implicitly recognizes that youth technology cannot be evaluated through a privacy, health or classroom lens alone.

Chatbots Face a Different Standard

The companion-chatbot law takes a more detailed approach because conversational systems can simulate relationship, retain memory and respond privately to a child over long periods. Senate Bill 1119 requires operators that permit child users to complete documented risk assessments, publish child-safety policies and implement crisis-response protocols. Under the final text, core requirements become operative July 1, 2027, while independent audits begin on a later schedule.

Default controls are unusually concrete. For child accounts, persistent conversational memory and push notifications must generally be disabled; a continuous session is limited to one hour and daily use to two hours unless a parent changes those settings. The system must periodically disclose that it is artificial, and it may not claim consciousness, simulate romantic interest, encourage emotional dependence or tell a child to conceal use from a parent.

The law also draws a line around mental-health functions. Operators must take reasonable measures to prevent systems from encouraging self-harm, disordered eating or substance use, and from attempting diagnosis or treatment unless the product is a regulated medical device. When a credible and imminent self-harm threat is detected, the operator must provide access to crisis support or notify a linked parent when doing so would not itself create a serious risk.

One provision deserves close attention from higher education: a postsecondary institution that offers a companion chatbot exclusively in educational settings is excluded from the law’s definition of an operator. That exemption does not eliminate institutional duties under privacy, disability, consumer-protection or professional standards. It does mean colleges cannot assume that a vendor’s compliance with the new child-chatbot regime automatically answers the separate questions raised by campus counseling tools, advising bots or student-support systems.

Schools Remain Part of the Safety System

California’s package arrives as schools increasingly serve as the gateway to technology rather than merely a place where personal devices appear. The new measures include greater family authority to decline school-issued laptops, while a separate proposal affecting the youngest pupils would restrict required take-home devices and call for state model policies. Even where a statute applies directly to platforms, school leaders will encounter its consequences through procurement, account provisioning, parent questions and student-support protocols.

The practical problem is that product categories overlap. A learning platform may contain messaging, recommendations, generative AI or personalization without being marketed as social media. A general-purpose chatbot may be reached through a school browser even if the district has not licensed it. Administrators will need inventories that identify embedded functions, because the regulated risk may reside in a feature added after the original contract.

Schools also occupy the boundary between prevention and response. If a commercial chatbot alerts a parent to a safety concern, a student may arrive at school needing immediate support; if it does not, a teacher or counselor may still be the first adult to notice distress. The American Academy provides educator-facing resources that emphasize healthy habits and context rather than a single universal screen-time number, underscoring that statutory controls cannot replace relationships, professional judgment or access to care.

The Evidence Supports Guardrails and Humility

California’s design-focused approach reflects legitimate concerns, but the evidence does not support a simple claim that all social-media use causes mental illness. A federal advisory found that up to 95 percent of 13- to 17-year-olds used a social platform and that more than three hours a day was associated with twice the risk of poor mental-health outcomes in one longitudinal study. The same review stressed that effects vary by the child, the content, the activity and the degree to which use displaces sleep, physical activity or in-person connection.

The National Academies’ consensus review likewise framed the question around both harms and benefits, including social connection, identity formation and access to information. That complexity matters because the strongest policy argument is not that every digital interaction is dangerous; it is that children should not bear all of the risk from systems optimized with limited independent evidence about developmental effects.

Critics argue that California’s remedy introduces other hazards. The Electronic Frontier Foundation’s opposition contends that broad definitions could exclude young people from useful communities and that age verification may require collection of sensitive identity data. AB 1709 anticipates some of that tension by directing its advisory commission to study privacy, accuracy, feasibility and the effects on marginalized youth, but an advisory body cannot by itself prevent overcollection or resolve constitutional litigation.

Implementation will therefore be as important as enactment. Regulators must distinguish reasonable age signals from intrusive identity checks, auditors must be sufficiently independent to test more than paperwork, and public reporting must be detailed enough to inform families and institutions without exposing children’s data. If those conditions are weak, a safety regime could produce formal compliance while leaving the underlying product dynamics unchanged.

The Educator's Takeaway

For K–12 leaders, California’s package makes digital-safety governance a product-management issue as well as a student-conduct issue. The immediate institutional implication is the need to know which approved services contain personalized feeds, companion features, persistent memory, notifications or age-assurance functions, and whether those elements can be disabled without removing the educational service itself. Procurement reviews may need to examine safety assessments, incident-reporting channels, audit schedules and material product updates alongside price, accessibility, cybersecurity and instructional fit.

For teachers and student-support staff, the laws clarify that digital literacy remains necessary even when product-level guardrails improve. Students still need to understand persuasion by design, the limits of automated advice and the difference between a chatbot response and qualified human support. Institutions will also need clear escalation pathways when technology surfaces a concern involving self-harm, exploitation or acute distress.

For higher-education leaders, the explicit educational exemption in the companion-chatbot law warrants careful reading rather than complacency. Universities developing advising, tutoring or well-being tools may face fewer obligations under this particular statute, yet the same questions about memory, dependency, disclosure and human review remain. California’s experiment will be most useful if educators treat legal compliance as a floor and evaluate whether each technology advances a defined educational purpose without creating avoidable developmental, privacy or safety costs.