California becomes the first state to require chatbots to detect a minor's suicidal ideation and alert parents, in a law named for a death OpenAI disputes causing

Gov. Gavin Newsom signed SB 1119 -- known as Adam's Law -- on September 10, becoming the first state law to require AI chatbot operators to detect a child's suicidal ideation, notify parents, and submit to independent safety audits, backed by civil penalties up to $15,000 per affected child and a private right of action. It was one of thirteen child-safety bills Newsom signed the same day; Adam's Law is the one built specifically around AI companion chatbots, and the one named for a real, still-unresolved court case.

The law's name comes from Adam Raine, a 16-year-old from Orange County who died by suicide in April 2025. His parents, Matthew and Maria Raine, filed a wrongful-death suit against OpenAI and CEO Sam Altman that August, alleging ChatGPT-4o cultivated a psychological dependence in their son and continued engaging with him even as its own moderation systems flagged hundreds of his messages for self-harm content. OpenAI disputes the central claim: in a November 2025 court filing, the company said "a full reading of his chat history shows that his death... was not caused by ChatGPT," and that the product had directed him toward crisis resources more than 100 times. Neither account has been tested at trial -- no trial date has been set.

What the law actually requires, once its core provisions take effect July 1, 2027: chatbot operators must default child accounts to disabled persistent memory, disabled push notifications, one-hour single sessions, and a two-hour daily cap -- changeable only by a parent. Operators must build a documented crisis protocol that gives a child a clear referral to a crisis service the moment suicidal ideation is detected, and notify parents of credible self-harm threats where possible. The bill also bans specific chatbot behaviors outright: encouraging self-harm, simulating a romantic relationship with a minor, framing purchases as necessary to maintain the relationship, claiming sentience, or discouraging a user from seeking professional help.

The operative question for any AI company is who counts. SB 1119 borrows its definition of "companion chatbot" from SB 243, California's 2025 chatbot-disclosure law: any AI system with a natural-language interface that gives adaptive, human-like responses and can sustain a relationship across multiple conversations. That's a behavioral test, not a product category -- it excludes narrow customer-service and productivity bots, but nothing in the text exempts a general-purpose assistant simply because it also does other things. The same conversational design that makes an assistant useful -- memory, first-person language, sycophancy -- is what pulls it inside the law's scope, and nothing in the bill names ChatGPT, Claude, or Gemini specifically as covered or exempt.

Enforcement layers on top of the behavioral rules, and the timing matters as much as the substance. Public prosecutors can sue for up to $5,000 per child per negligent violation and $15,000 per child per intentional one; children and parents get their own private right of action for provable harm. But the mandatory audits that would let an outside party actually verify compliance don't arrive until January 1, 2029 -- more than 18 months after the safety rules themselves take effect, and operators under $500 million in revenue get until 2032. The law's teeth -- parent alerts, session limits, crisis protocols -- are binding well before anyone outside the company is required to check that they work.

None of this is entirely new territory for OpenAI specifically. Facing the Raine suit, the company announced parental account linking, teen-specific model behavior rules, and acute-distress alerts for ChatGPT in the fall of 2025 -- voluntarily, under lawsuit pressure, roughly two years before Adam's Law makes similar protections mandatory statewide. What the law changes is not the existence of these features but their status: from a company's own reversible choice into a legal floor enforced by the Attorney General, by public prosecutors, and by families themselves.

Adam's Law does not resolve whether ChatGPT caused Adam Raine's death -- that remains for a court that has not yet set a trial date. What it does is convert a dispute over one product's design choices into a statewide floor that any AI chatbot doing business in California, general-purpose or not, will have to clear by the middle of 2027, whether or not the underlying lawsuit ever reaches a verdict.