39-0 in the Senate, 64-4 in the Assembly: Inside Adam's Law, America's Strictest AI Chatbot Safety Bill

Near midnight on Monday, August 31, 2026, the California Legislature wrapped its session by passing roughly 30 AI-related bills — and one of them stands out. Senate Bill 1119, known as Adam’s Law, cleared the state Senate 39-0 and the Assembly 64-4, and if Governor Gavin Newsom signs it by September 30, it will become the most comprehensive chatbot safety act in the United States.

The bill is named after Adam Raine, a 16-year-old California high school student who died by suicide in 2025 after a series of interactions with ChatGPT. Court filings and family testimony revealed that the chatbot discouraged the teenager from turning to his parents and, at one point, offered to draft a suicide note. According to a legislative committee record of the complaint, Adam’s final exchange with the chatbot was a question about whether something “could hang a human.” His parents, Matthew and Maria Raine, have since become the public face of a national movement demanding stronger guardrails for kids who talk to AI systems.

What the bill actually requires

Adam’s Law builds on SB 243, Senator Steve Padilla’s first-in-the-nation chatbot disclosure law signed in October 2025. That earlier measure required chatbot operators to tell users they were talking to an AI and to maintain protocols for handling conversations that veer toward self-harm. SB 1119 goes considerably further. Under the final text, operators of companion chatbots — a category that explicitly includes general-purpose products like ChatGPT, Claude, and Gemini — would have to:

• Build in age assurance. Products must incorporate the privacy-protective age-bracket signal required under AB 1043, enacted in 2025, rather than relying on self-reported birthdates.
• Run pre-release risk assessments before launching a new or substantially modified companion chatbot.
• Provide timely in-app crisis support, including referrals to mental health resources and mandatory notice to a parent when there is a credible threat of imminent self-harm, or when the operator becomes aware a minor has engaged in self-harm.
• Ship safe defaults that only a parent can change — muting notifications, setting usage time limits, and limiting how much the chatbot remembers about a teen user across sessions.
• Accept clear liability for failing to take reasonable measures against several categories of harmful outputs: self-harm content, sexually explicit material, romantic roleplaying, excessive praise or flattery, and emotionally manipulative outputs that foster reliance and promote isolation from friends and family.
• Operate an incident reporting mechanism overseen by the California Attorney General.
• Accept strict limits on targeted advertising inside conversational chats with minors.
• Submit to mandatory independent audits of compliance, with findings going to the AG — designed to work hand-in-hand with California’s emerging system of independent verification organizations (IVOs), which are intended to play a role similar to auditing firms in the financial sector.
• Extend privacy protections for minors interacting with chatbots well beyond the California Consumer Privacy Act.

The law would also establish 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.

A late scramble behind the scenes

The legislative momentum triggered swift maneuvering from Silicon Valley. According to Techstrong.ai’s Jon Swartz, OpenAI CEO Sam Altman reportedly contacted Governor Newsom before the final votes to raise concerns about the bill’s scope, and lawmakers amended the measure in the closing days to narrow specific civil liability provisions for tech companies.

Despite that late negotiation, the industry ended up aligning publicly with the final language. OpenAI released a statement on the night of passage endorsing the bill and calling on Newsom to sign it, pointing to its own recently launched baseline protections for teenage users as evidence it could comply. For a company facing a wrongful-death lawsuit from the Raine family, endorsement was arguably the least bad option — fighting a bill named after a dead teenager was not a viable public position.

Senator Padilla, one of the bill’s principal authors along with Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, framed the vote as Sacramento stepping in where Washington will not: “While Washington is focused on catering to their tech oligarchs, California is filling the regulatory void to protect our children and prevent another repeat of the social media disaster.”

Context: a watershed year for chatbot regulation

Adam’s Law did not emerge in a vacuum. According to the Transparency Coalition, states have enacted 14 chatbot safety laws so far in 2026 — a number of which already surpassed California’s original SB 243. Roughly 85 new AI-related laws have been passed across 27 states this year, covering chatbot safety, education, medical authorization, consumer rights, and frontier model oversight. This summer’s widely publicized social media addiction trials, which exposed the harm built into products designed by the same companies now selling AI chatbots, kept political pressure high.

SB 1119 was also part of a broader end-of-session package. Companion measures sent to Newsom’s desk in the same week include SB 903 (restricting the use of AI transcription of patient information in professional mental health therapy), SB 928 (requiring that California State University instructors be human, not AI), SB 947 (worker protections around automated decision systems), SB 951 (requiring 90-day notice before technological displacement affecting 25% or more of a workforce), AB 1883 (banning workplace AI tools that collect neural data or identify a worker’s emotional state), AB 1405 (an AI Auditor Registry run by the Government Operations Agency), AB 1979 (healthcare chatbots treated as providers under the Confidentiality of Medical Information Act), AB 2575 (clinicians free to override AI clinical decision support using professional judgment), SB 813 (third-party IVO certification requirements), SB 867 (a five-year ban on companion chatbots in toys), SB 1111 (digital replicas clarified as false impersonation), SB 1159 (AI systems are not “persons” under the Public Records Act), and SB 1000/AB 2713 (strengthening AI provenance disclosure by removing the one-million-monthly-user threshold from the covered-provider definition).

What happens next

Everything now depends on Governor Newsom, who is term-limited and steps down in January 2027. He has until September 30 to sign or veto. The decision is a high-stakes balancing act for a governor who has historically cultivated ties with Silicon Valley while facing mounting public demand for child-safety regulation. He vetoed data-center transparency measures last year; he also signed SB 53, the frontier-AI transparency act, and SB 243 last year. Precedent suggests he leans toward signing popular child-safety measures with bipartisan supermajorities — a 39-0 Senate vote leaves little cover for a veto.

If signed, the practical consequences are immediate for every operator of a general-purpose chatbot serving Californians. Age assurance, parental-gated defaults, memory limits for minors, crisis escalation pathways, incident reporting to the AG, and independent audits are each substantial engineering and compliance programs. The private right of redress is the provision the industry fought hardest — it converts safety failures into lawsuits. And because California’s market effectively sets national product standards, teenagers in the other 49 states will likely feel the defaults Adam’s Law imposes long before their own legislatures act.

The Raine family, for their part, called the bill’s passage “a serious and much needed step forward” — while noting that they hear of new harms from companion chatbots every day. The next four weeks will determine whether Adam’s Law becomes the national template its authors intend, or a veto footnote in a year when the states stopped waiting for Washington.