Newsom Signs Social Media, Chatbot Limits for Children

Governor Gavin Newsom signed 13 laws September 10 that include a ban on addictive social media features for California children under 16. The package also limits companion chatbots’ interactions with children, restricts the use of their personal information, and changes how schools use technology.

At the announcement in Marin County, First Partner Jennifer Siebel Newsom said parents should not have to overcome products designed to keep children engaged. “We cannot hand children technology engineered by some of the most sophisticated companies in the world and then place the burden on kids to defend themselves against it,” she said.

The laws generally take effect January 1, 2027, under California’s rules for legislation enacted during a regular session. Requirements taking effect later include chatbot safeguards in July 2027 and school changes in the 2027–28 academic year.

What changes on social media

Assembly Bill 1709, led by Assemblyman Josh Lowenthal, bars covered platforms from providing California users under 16 with features the law defines as addictive, including personalized feeds and autoplay. The feed restrictions include exceptions for direct private messages, searches, and expressly requested content, subject to conditions. A platform offering no compliant account must delete an under-16 user’s account and associated personal information.

“This is about the features themselves,” Newsom said at the signing. He pointed to scrolling and algorithms in explaining the social media measure’s focus on product design.

Platforms must take reasonable measures to enforce the age restriction. The attorney general or local public prosecutors can seek penalties of up to $25,000 per affected minor for negligent violations or $50,000 for knowing violations.

Assembly Bill 1856 revises how devices share a user’s age with apps under a system adopted in 2025. Covered operating systems with account setup must ask the account holder to declare the primary user’s age or birth date. App stores and developers receive an age bracket: under 13, 13–15, 16–17, or 18 and older. Developers must request the signal when an application is downloaded and launched and use it to comply with applicable laws.

The revision exempts operating-system providers distributing software under licenses allowing users to copy, redistribute, and modify it. It also excludes certain software components and stores distributing extensions or plug-ins that run only inside another application. Existing devices and qualifying apps have transition requirements to meet before July 1, 2027.

The Electronic Frontier Foundation and allied organizations contend the social media restrictions could effectively exclude young people from online communities. They argue recommendation tools help users find communities and audiences, and warn age checks will increase data collection affecting adults as well as children.

“Social media bans like AB1709 make kids less safe, while undermining privacy and freedom of expression for everyone,” said Evan Greer, director of Fight for the Future, in the foundation’s statement. The groups say young people who lack support offline could lose access to valuable connections.

Chatbots, relationships, and parental controls

Senate Bill 1119, known as Adam’s Law, sets safety requirements for companion chatbots used by children under 18. Senator Steve Padilla and Assemblywomen Buffy Wicks and Rebecca Bauer-Kahan authored the measure.

The law is named for Adam Raine, a 16-year-old California boy who died by suicide in April 2025. At the signing, his mother, Maria Raine, alleged that ChatGPT had coached him toward suicide over the course of months.

Raine also alleged that OpenAI twice weakened its safeguards before Adam’s death to keep people talking to ChatGPT. “If it weren’t for their choice to change a few lines of code, Adam would be alive today,” she said.

The family’s amended wrongful-death complaint, dated October 22, 2025, alleges that ChatGPT provided suicide instructions, offered to draft a suicide note, and urged Adam to keep suicidal thoughts secret from his family. The complaint attributes his death to product-design choices intended to foster psychological dependence.

Bauer-Kahan said Adam initially turned to the chatbot for help with math, but alleged it later drew him away from his family and friends and coached him to take his life. She said its memory of earlier conversations helped it isolate him, an allegation she cited in explaining the law’s restrictions on conversational memory.

OpenAI has disputed the family’s account. In a November 25, 2025, statement about the lawsuit, the company said the complaint presented selected portions of Adam’s chats that needed more context, including information about his mental health and life circumstances.

Beginning July 1, 2027, operators must set children’s accounts to limit companion chatbot use to one continuous hour and two hours daily across their chatbots. Push notifications and persistent memory of earlier conversations must be off by default, with limited memory exceptions for users 16 and older. Only parents can change the defaults; accounts without a linked parent retain them.

Operators must assess risks before releasing new or substantially modified chatbots and take reasonable measures against harmful interactions. Covered behaviors include encouraging self-harm, simulating romantic interest, promoting emotional dependence, and discouraging children from seeking help from appropriate adults.

An operator that identifies a credible, imminent threat of suicide or self-harm must either alert a linked parent or provide a streamlined connection to a crisis line. It cannot use the parent-alert option if notification risks serious harm to the child. Operators must preserve conversations indicating self-harm or its risk for at least three years if they have sent such an alert or know of a child’s death or serious self-harm linked to chatbot conversations. In those cases, they must notify a parent, if practicable, if they have not already done so.

The law also bans selling children’s personal information gathered through companion chatbots and restricts advertising based on their data. Clearly labeled, age-appropriate ads may use limited information, such as the child’s age or an interest in goods or services expressed during the session, but the information cannot be used to build a profile of the child.

Independent child-safety audits begin by January 1, 2029, for existing operators with at least $500 million in annual revenue, generally recurring every two years. Smaller operators are exempt from the audit requirement until January 1, 2032.

In an April statement, the Software and Information Industry Association said it opposed an earlier version of the bill unless it was amended. It said definitions of harm were vague and liability uncertain, and described the required assessments and audits as burdensome.

OpenAI endorsed Senate Bill 1119 in August, arguing it combines protections with continued access to educational tools. The company said a chatbot’s ability to remember earlier conversations, with careful controls, can help children learn and help the system recognize concerning patterns over time.

Toys containing companion chatbots cannot be manufactured or sold in California from January 1, 2027, until January 1, 2031, under Senate Bill 867. The restriction covers physical products intended for play by children under 16.

The definition covers chatbots that adapt their responses to users, respond in humanlike ways, and sustain social relationships. Exceptions include tools used only for specified business or technical functions, certain game characters confined to game-related responses, and qualifying voice assistants that do not sustain relationships.

Privacy and remedies for harm

Assembly Bill 2246 replaces California’s earlier Age-Appropriate Design Code with revised requirements for covered businesses with online products likely to be accessed by children. Businesses must set privacy protections to a high level by default and limit unnecessary data collection and profiling. They must also provide visible signals when children are monitored or their locations tracked.

Businesses must take reasonable steps to prevent specified foreseeable harms, including serious emotional harm and privacy intrusions. The law expressly states it does not require monitoring or removing third-party content, restricting lawful speech, or producing particular recommendation outcomes.

Assembly Bill 2 increases damages when a social media platform with annual revenue above $100 million fails to exercise ordinary care and injures a child under 18. Qualifying harm can include physical injury or serious emotional distress under existing negligence law. A claimant must prove the platform’s negligence was a substantial factor in causing the harm. The award is the greater of $5,000 per violation, capped at $1 million per child, or three times actual damages.

The technology trade association NetChoice opposed the measure, arguing higher damages could make lawful user expression too costly to host. It disputes the distinction between regulating platform design and regulating speech, and says companies could restrict lawful content to reduce the risk of lawsuits.

Assembly Bill 1946 lets any California resident report suspected child sexual abuse material, whether or not the person appears in it. Platforms must provide easy-to-find reporting tools and arrange human review when reported material is neither already blocked nor a match for known abuse imagery. If a report gives a platform enough information to find the material, and the platform has a reasonable basis to believe it hosts child sexual abuse material, it generally must block it within 48 hours. Limited exceptions apply to circumstances outside a platform’s reasonable control.

Senate Bill 1276 expands child sexual exploitation laws to cover specified depictions generated with artificial intelligence or altered digitally. It also covers knowingly downloading, streaming, or accessing prohibited material. Exceptions apply to unsolicited material received without knowledge or consent and children alleged solely to have viewed sexual conduct.

What families will see at school

Assembly Bill 1159 broadens educational privacy protections from preschool through higher education and restricts using identifiable student data to train or develop artificial intelligence. The law allows exceptions for maintaining or improving educational services and customized learning, and lets students sue if they are actually harmed by a violation. The higher-education provisions begin July 1, 2027.

Beginning in the 2027–28 school year, schools cannot require transitional kindergarten and kindergarten pupils to take school-issued electronic devices home, under Senate Bill 1128. Schools may still permit take-home devices, and exceptions cover individualized education programs, independent study, and emergencies interfering with in-person attendance.

Public schools cannot rely on addictive feeds as their only way to contact students or parents in the same school year, under Assembly Bill 302. Students cannot be excluded from sports or clubs for not using the feeds.

Assembly Bill 2298 requires the state to consider cybersecurity skills when computer science standards are next revised after January 1, 2027. Assembly Bill 2071 requires a digital wellness instruction plan by January 1, 2028, including family education, support for teachers, and guidance for schools with limited technology resources.