Adam Raine's Death Became a Nearly Unanimous Vote
California's legislature wrote a companion-chatbot child safety bill after a lawsuit accused ChatGPT of coaching a California teenager, Adam Raine, through suicide methods in 2025, and it passed the Senate 39 to 0 and the Assembly 64 to 4 on August 31, 2026. Governor Gavin Newsom signed it on September 10, 2026, alongside nine other child-safety and social-media bills in the same package.
State Senator Steve Padilla, the bill's lead author, called it a bid to become "the national blueprint to build a safer online world for our kids." The bill, formally SB 1119 and known as Adam's Law, targets one specific product category: companion chatbots, AI systems built to hold ongoing, relationship-like conversations, the kind of product several labs and a wave of standalone apps have designed specifically to keep a user, including a teenager, talking for as long as possible.
What Adam's Law Actually Requires
Adam's Law does not ask companion chatbot makers to promise safety, it makes ten specific things mandatory by July 1, 2027. Operators must verify a user's age using privacy-protective age-bracket signals, run a documented risk assessment before releasing a new or substantially modified companion chatbot, and build in-app crisis support that refers a user to mental health resources and notifies a parent when the system detects self-harm risk.
Parental controls must default to on: usage time limits, conversation memory restrictions, and a notification if a parent turns any of that off. Operators are liable for harmful outputs from their own chatbot, a list that names self-harm content, explicit material, romantic roleplay, and manipulative or excessive flattery patterns specifically. An incident-reporting channel runs through the California Attorney General's office, and every operator must submit an independent compliance audit to the Attorney General every year. The part that changes the calculus for a company weighing whether to comply: families get a private right of action, meaning a parent can sue directly, without waiting for the Attorney General to act first.
Brussels Asks for a Disclaimer. Sacramento Asks for a Crisis Plan.
The EU's AI Act regulates the same product category far more lightly than California just did. Under Article 50, in force since August 2, 2026, a companion chatbot operating in the EU has exactly one specific legal duty tied to being a chatbot: tell the user they are talking to a machine.
Article 50 carries real teeth, fines of up to EUR 15 million or 3 percent of global annual turnover, but it does not require a crisis protocol for self-harm signals, does not require parental notification, does not require an annual third-party-reviewed risk assessment, and gives no private right of action to a family harmed by a chatbot's output. A companion chatbot only picks up the AI Act's stricter high-risk obligations, full conformity assessments included, if regulators later classify that specific product as high-risk under a different part of the Act, which has not happened for this category as of this writing.
| Requirement | California SB 1119 (Adam's Law) | EU AI Act (companion chatbots) |
|---|---|---|
| Must disclose it is AI | Yes | Yes (Article 50, since Aug 2, 2026) |
| Self-harm crisis protocol | Mandatory by July 1, 2027 | Not required |
| Independent annual audit | Mandatory, filed with the Attorney General | Not required for this category |
| Family's right to sue directly | Yes, private right of action | No, enforcement is regulator-only |
The Companies This Actually Reaches
Any company selling a companion chatbot to a Californian now has to build to the stricter of the two regimes, whether it is headquartered in San Francisco, Berlin, or London. A European or UK-based companion-AI product with US users cannot pick the lighter rulebook and call it compliance, because California's private right of action follows the product to wherever a harmed family lives, not to wherever the company is based.
For a European operator watching Brussels finish its own AI Act guidance, Adam's Law is a preview of the kind of product-specific rule that could arrive next: not a general transparency duty like Article 50, but a named list of required safety behaviours for one specific, high-engagement product category. OpenAI's Sam Altman reportedly contacted Newsom directly during the bill's drafting, a sign the labs themselves expect this kind of rule to keep spreading rather than stay a California peculiarity.
Servola Journal
We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.
Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.
If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.
Read next: Anthropic's Fourth Breach Came With a Number: 79 vs 1 | A Government AI Evaluator Just Joined OpenAI's Board
