Adam's Law (SB 1119): What California Now Requires of AI Companion Chatbots

Governor Newsom signed Senate Bill 1119, known as Adam's Law, on September 10, 2026. Its authors, Senator Steve Padilla and Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, describe it as the first comprehensive regulatory framework in the country for AI companion chatbots and children.

The bill is named for Adam Raine, a California teenager who died in 2025. His parents have filed a wrongful death lawsuit against OpenAI alleging that Adam spent months confiding in ChatGPT before his death, and his father has testified before Congress that a tool the family understood as a homework helper became his son's closest confidant. OpenAI disputes the allegations, and the case remains in litigation.

The Legislature's broader conclusion didn't depend on how that case comes out. Products designed to form emotional bonds with children had been operating with almost no specific rules, and SB 1119 is California's attempt to write some.

What operators have to do

According to the authors' summary, the law requires operators of covered companion chatbots to put in place:

• age assurance, using the privacy-protective age-bracket signal established by AB 1043 (Wicks, 2025)
• risk assessments before releasing a new or substantially modified companion chatbot
• timely in-app crisis support, including referrals to mental health resources, and notice to a parent where there's a credible threat of imminent self-harm or the operator is aware the child engaged in self-harm
• protective default settings that only a parent can change, including no push notifications, usage time limits, and limits on persistent conversational memory
• reasonable measures to prevent specified harmful outputs: self-harm content, sexually explicit material, romantic roleplay, excessive praise or flattery, and emotionally manipulative responses that foster reliance and isolate a child from friends and family
• an incident reporting mechanism overseen by the Attorney General
• strict limits on targeted advertising to children within chats
• privacy protections the authors describe as going beyond the California Consumer Privacy Act
• mandatory independent compliance audits, with findings submitted to the Attorney General

The default limits

The bill text is specific about time. By default, an operator must limit a child to one hour in a single continuous session and two hours a day in total across the companion chatbots it controls. Parental controls must let a parent adjust each default. Persistent conversational memory can't be provided to a child by default unless the operator puts in place specified guardrails against the degradation of its child-safety measures, so it's a conditional restriction rather than a flat ban. Beginning January 1, 2027, operators must either determine users' ages or apply the child-safety protections to everyone.

Families can bring their own claims

For families, the most consequential provision may be the enforcement one. Many consumer protection statutes can only be enforced by government agencies, which leaves the people actually harmed without a way to sue. According to the authors, Adam's Law includes a private right of redress for families harmed by violations relating to crisis referral services, parental notice, default protections, and the prohibitions on harmful outputs.

In practical terms, a family doesn't have to wait for the Attorney General. If an operator failed to provide crisis referrals when a child expressed suicidal thoughts, failed to notify a parent when the law required it, stripped away the default protections, or failed to take reasonable measures against the harmful outputs the statute lists, the family itself may be able to pursue a claim.

What counts as a companion chatbot

The law is aimed at AI systems built to feel like a relationship: products that present as a friend, remember past conversations, send notifications pulling a child back in, and respond with affection and validation. An ordinary search tool isn't the target.

Notice that the prohibited behaviors, like encouraging emotional dependence, flattery out of proportion to the situation, or helping a child get around parental controls, are about how the product is designed to behave, not about anything the user types.

How it fits with earlier laws

The authors describe SB 1119 as building on three 2025 measures: SB 243 (Padilla), which set baseline disclosure and self-harm detection rules for companion chatbots; AB 1064 (Bauer-Kahan), whose "safe-by-design" approach it restores; and AB 1043 (Wicks), which requires privacy-protective age signals at the operating-system level.

The bill passed the Senate unanimously and the Assembly 64 to 4, and OpenAI publicly supported it. That kind of consensus is unusual for technology regulation in California. Several requirements phase in during 2027, though, and age verification mandates in particular have drawn concern from technology and civil liberties groups because of the sensitive information they can require companies to collect. A law this broad may face legal challenges, and how courts will treat its provisions isn't yet known.

For parents

Find out which apps your child uses, and whether any of them are companion-style AI products rather than ordinary assistants. Once the defaults take effect, notifications, time limits, and memory should be set protectively unless a parent changed them, and if they aren't, that's worth noting. Talk with your child about how much time they spend with these products and whether one has started to take the place of people.

If something concerning happens, preserve the conversations right away with screenshots or an export. Chat histories can be deleted by the child or by the platform. If your child expressed distress and the product didn't point to crisis resources, or you were never notified when you should have been, the timing and content of those exchanges will be at the center of any claim.