On September 10, 2026, California Governor Gavin Newsom signed Senate Bill 1119 — Adam's Law — establishing what its authors describe as the most comprehensive framework in the United States governing how AI companion chatbots may interact with children. The bill passed the Senate unanimously and cleared the Assembly 64-4 in late August, arriving on the governor's desk as part of a broader bipartisan package on child online safety.

The law is named for Adam Raine, a 16-year-old Californian whose parents, Matt and Maria Raine, sued OpenAI in San Francisco Superior Court on August 26, 2025. Both attended the signing ceremony and called the measure a well-considered and significant step, saying they were glad that part of their son's legacy is making chatbots safer for minors.

What SB 1119 Actually Requires

Adam's Law converts a set of voluntary industry practices into statutory obligations. The core duties reported in the legislation:

  • Pre-release risk assessment. Companies must assess a companion chatbot before releasing it, and again before any substantial modification. This is the provision with the broadest operational reach — it makes shipping a model update a documented compliance event, not just an engineering one.
  • Crisis referral and parental notification. Platforms must refer users to mental health resources where appropriate, and notify a parent when there is a credible threat of imminent self-harm.
  • Parental controls on by default. Limits on notifications, usage and persistent conversational memory must be enabled by default and changeable only by a parent. Default-on memory limits are a notable design mandate, given how central persistent memory has become to companion products.
  • Notification when safety settings are disabled, so a child cannot silently switch protections off.
  • Liability for harmful outputs, where a provider fails to take reasonable measures to prevent them.

The wider package that accompanied it bans addictive design features — autoplay and algorithmic feeds built on a user's history and profile — for users under 16, restricts targeted advertising to children, expands children's privacy protections, and provides for penalties of up to $1 million per child against large social media companies found negligent in harming children. A separate measure, SB 867, imposes a moratorium on AI toys with companion chatbot functionality.

The bills were authored by Senator Steve Padilla (D–San Diego) and Assemblymembers Buffy Wicks (D–Oakland) and Rebecca Bauer-Kahan (D–Orinda). Newsom framed the signing plainly: as innovation moves faster, protections must keep pace, and innovation comes with responsibility. Padilla's framing was more pointed — California, he said, is filling the void created by inaction in Washington.

Why It Matters

Adam's Law is the first American statute to impose product-safety-style duties on conversational AI aimed at minors, and the mechanism it chose matters more than the headline. By requiring a documented risk assessment before release and before substantial modification, the law regulates the development process, not just outcomes. That is the same regulatory architecture the EU AI Act uses, and it travels: a company that builds assessment tooling to satisfy California has most of what Brussels asks for.

Three practical consequences for anyone shipping conversational AI:

  • Compliance is now a release gate. Teams practising continuous deployment on consumer chatbots need a documented, repeatable assessment they can produce on demand — which in turn requires versioned records of model, system prompt and safety configuration.
  • Age signal becomes load-bearing. Almost every duty in the law is conditioned on the user being a minor, which pushes age assurance from a policy checkbox to core infrastructure.
  • Memory becomes a regulated feature. Persistent conversational memory — the thing that makes companion products sticky — is now subject to a default-off-by-parent regime for minors in the largest US state market.

The Legal Vulnerability

California has been here before, and lost. SB 976, the 2024 measure targeting addictive feeds, was blocked by the Ninth Circuit on January 28, 2025 after NetChoice sued on First Amendment grounds. That history explains the drafting of this package, which leans on strict age verification and negligence-style penalties rather than direct restrictions on speech or content — an attempt to build on firmer constitutional ground.

Whether it holds is an open question, and a challenge should be considered likely. Compelled disclosures, mandated default settings and liability for a model's outputs each offer a line of attack. The parental-notification requirement in particular sits in uncomfortable territory, balancing minors' privacy interests against parental authority and a platform's judgement about what constitutes a credible threat.

Where This Sits in the Wider Regime

Adam's Law builds on a fast-accumulating California stack: SB 243, signed in October 2025, was the first state law requiring AI chatbot operators to implement companion safety protocols, and TFAIA, signed in September 2025, made California the first state with a statute specifically addressing frontier model development, with whistleblower protections and published safety test results.

Other states are moving on parallel tracks — Texas under TRAIGA with an attorney general complaint mechanism, Colorado's ADMT Act and Chatbot Safety Act, and Connecticut's CART Act with frontier-developer duties phasing in from late 2026. Enforcement is arriving too: the Kentucky Attorney General sued Character Technologies and its founders in January 2026 over the Character.AI product.

Running against all of it is Executive Order 14365, signed in December 2025, which establishes a federal policy addressing state AI laws and reflects the administration's push for Congress to pre-empt this exact patchwork. That collision — a state writing detailed product duties while the federal government argues for a single national framework — is the defining structural fight in US AI policy, and Adam's Law just raised its stakes considerably.

Sources