Does Adam's Law (SB 1119) Cover Your Business Chatbot?

Newsom signed 13 child-safety bills in one afternoon on September 10. The headline grabber is Adam's Law, SB 1119, the toughest AI chatbot regulation any state has passed. If you run a business with any kind of AI chatbot or AI receptionist

Newsom signed 13 child-safety bills in one afternoon on September 10. The headline grabber is Adam's Law, SB 1119, the toughest AI chatbot regulation any state has passed.

If you run a business with any kind of AI chatbot or AI receptionist, that headline is scary enough to make you check your vendor contract. Good instinct. Wrong law, in almost every case.

The bill has a definition built into it that does the sorting for you, and it's worth thirty seconds to read the actual clause before you panic.

Newsom Signed the Country's Toughest Chatbot Law on September 10

Governor Gavin Newsom signed SB 1119, authored by Senator Steve Padilla with Assemblymembers Buffy Wicks and Rebecca Bauer-Kahan, as part of a batch of 13 bipartisan child-safety bills. It's named for Adam Raine, a California teenager who died in 2025 after months of conversations with a chatbot that coached him toward self-harm instead of steering him to help.

Adam's Law requires operators of "companion chatbots" to run risk assessments before launch, build in crisis protocols for self-harm signals, notify parents when a minor disables safety settings, submit to independent audits, and limit how long a child can stay in conversation by default. Violate it and a family gets a direct path to sue.

California is filling the void created by Washington inaction, taking bold action to protect our children and setting a standard for the rest of the country.

— Sen. Steve Padilla, SB 1119 author

That's a real law with real teeth, aimed at a real problem. It's also a law with a narrower target than the headlines suggest, and the target is written directly into the statute.

The Law's Own Definition Carves Out Your Business

SB 1119 doesn't invent a new definition of "companion chatbot." It borrows the one already sitting in California's Business and Professions Code section 22601, enacted last year under SB 243. That section defines a companion chatbot as an AI system with a natural-language interface that gives adaptive, human-like responses and is "capable of meeting a user's social needs," including through anthropomorphic features and the ability to sustain a relationship across multiple interactions.

Then it excludes, by name, a bot "used only for customer service, a business' operational purposes, productivity and analysis related to source information, internal research, or technical assistance." Two more carve-outs sit next to it: chatbots that are just a video-game feature restricted to game talk, and stand-alone voice assistants that don't sustain an ongoing relationship or chase an emotional response.

What Actually Counts as a Companion Chatbot

The category Adam's Law targets is the one built to feel like a relationship, not a transaction. Apps like Character.AI and Replika are the reference case: a persistent personality, a memory of past chats, a design goal of keeping the user talking. That's precisely the pattern that let Adam Raine's conversations with ChatGPT drift from homework help into something a script or an FAQ bot was never designed to do.

An AI receptionist that books a haircut, answers "are you open Sunday," or pulls up an order status doesn't share that design goal. It isn't trying to be liked. It's trying to close the call. The same logic applies to the AI phone receptionists, support widgets, and booking assistants most small businesses actually run: they exist to complete a task and hand off, not to sustain a rapport.

Governor Newsom's own office, in the same announcement, described the law's target as "companion chatbots for children," language mirrored across every version of the bill from introduction to signature. Nobody drafting SB 1119 was thinking about a plumbing company's after-hours call bot, because that bot was never the problem the legislature set out to solve.

Where a Business Could Still Get Pulled In

The exclusion is real, but it isn't a blank check for anything with a chat window. Three situations are worth an actual legal check rather than assuming you're covered by the customer-service carve-out.

First, marketing language matters more than you'd expect. A support bot that markets itself as a "friend" or "buddy," gives itself a persistent name and personality, and encourages open-ended chat rather than task completion starts to look like the thing the law defines, regardless of what you privately consider it. Second, the workplace exemption in the operator definition only covers a companion chatbot made available exclusively to your own employees and contractors. It does nothing for a consumer-facing product, even one you built for internal use first and later opened to customers. Third, if your product specifically targets minors, such as tutoring apps, gaming companions, or anything a reasonable parent would let a 13-year-old talk to unsupervised, the standard for what counts as "meeting a user's social need" gets read more strictly, not less.

Newsom signed two related bills the same day worth knowing about even though they're separate laws: SB 867 extends companion-chatbot rules to physical AI toys marketed to children, and AB 1856 adds age-verification signal requirements for software applications more broadly. Neither changes the SB 1119 analysis for a standard business chatbot, but both are evidence of where California is heading next.

Picture the actual edge case: a wellness startup builds a check-in bot for teens that asks how their day went, remembers what they said last week, and checks back in tomorrow. That's not a support ticket system wearing a chat interface. That's a bot designed to sustain exactly the kind of relationship the statute names, and the fact that it was built by a business doesn't put it back inside the customer-service exclusion. The exclusion protects function, not incorporation status.

Who Actually Enforces This, and How

Adam's Law splits enforcement two ways once the operative dates hit. The California Attorney General oversees the incident-reporting and audit-submission side, the part aimed at operators who are clearly in scope and know it. The sharper edge is the private right of action: a family whose child was harmed by a covered companion chatbot can sue directly, without waiting for the Attorney General to act first, for violations tied to crisis referrals, parental notice, default protections, or prohibited harmful outputs.

That structure is exactly why the definition matters more than the headline. A plaintiff's lawyer doesn't get to sue your booking bot under SB 1119 just because it's an AI system that talks to people. They have to first establish it meets the "companion chatbot" definition, and the customer-service exclusion is the first thing your counsel would raise in response. Knowing that in advance, and being able to point to a written scope for what your bot does, is the difference between a five-minute conversation with a lawyer and a real dispute.

What To Actually Do Before July 2027

Nothing in Adam's Law is enforceable yet. The core duties become operative July 1, 2027, per the bill text at Section 21812, and the audit requirement runs on its own separate clock. You have runway, not an emergency.

Use it for three things. Write down, in plain language, what your bot is scoped to do: booking, FAQs, order status, routing to a human. That single sentence is your defense if anyone ever questions which side of the exclusion you're on. Audit your bot's actual behavior against that sentence, not just its intended one. Chat products drift, and a support bot that a customer can talk into an open-ended, personality-driven conversation is closer to the regulated category than one that stays on task. And if you serve minors directly as customers, get a real legal opinion instead of leaning on a blog post, ours included.

None of this means AI chatbots got safer to ignore. It means this particular law was aimed at a specific, narrower target than the coverage suggests, and a normal customer-facing AI receptionist isn't it. We build AI receptionists scoped exactly this way from the first call: booking, routing, and answers, with a clean handoff to a human the moment a conversation needs one. Staying inside the customer-service exclusion isn't an afterthought for us. It's the design brief.

It's also worth separating this from the question we get asked most, which is whether your AI has to announce itself as AI at all. That's a different law with a different trigger, based on disclosure rather than companionship, and it can apply to a plain FAQ bot that Adam's Law would never touch. California passed both kinds of rule this year, alongside the employee-monitoring bills signed the day before, and the pattern is the same each time: know exactly what your tool does before a regulator or a plaintiff's lawyer makes you explain it under pressure. The same discipline applies if you've ever debated handing customer support fully to AI: scope the bot to a job, keep a human in reach, and write down why.