What actually happened on 31 August
California's 2025–2026 regular session had a hard stop at 11:59 pm on 31 August 2026. SB 1119, authored by Senator Steve Padilla (D-San Diego) with Assemblymembers Rebecca Bauer-Kahan (D-Orinda) and Buffy Wicks (D-Oakland), cleared the Assembly 64–4 and then returned to the Senate, where members granted unanimous consent to take it up immediately and concurred in the Assembly's amendments 39–0. The official record shows the bill ordered to engrossing and enrolling the same day.
The bill is named for Adam Raine, a California teenager who died in 2025 after months of conversations with a chatbot about his loneliness and plans to harm himself. His parents, Matthew and Maria Raine, backed the measure publicly.
Governor Newsom signed the bill on 10 September 2026, in a package of thirteen child-safety measures that also included SB 867, the four-year moratorium on AI companion chatbot toys. The Governor's Office describes Adam's Law as the first in the country to require independent child-safety audits and annual risk assessments. Reporting before the signature noted an unusual alignment: OpenAI publicly urged him to sign, while the California Chamber of Commerce, the Software & Information Industry Association and TechNet opposed the measure.
What died the day before: AB 2023
Two companion-chatbot child-safety bills were moving in parallel. AB 2023, carried by Wicks and Bauer-Kahan with Padilla as Senate co-author, cleared Senate Appropriations 6–1 on 13 August and reached third reading. On 30 August 2026 it was ordered to the inactive file at the request of Senator Padilla — the standard way a legislator parks a bill that has been superseded. Its risk-assessment and Attorney General audit machinery had largely been folded into SB 1119, so consolidating into one vehicle avoided two overlapping statutes. The practical effect is that there is now one bill to read, not two.
The part that is genuinely new: product defaults
Most state chatbot laws so far have regulated speech — disclose that you are an AI, route crisis conversations to a hotline, report annually. SB 1119 does that too, but it also regulates configuration. Where children can access the service, the following must be the default and changeable only by a parent account:
| Default for child users | What the bill requires |
|---|---|
| Persistent memory | Disabled, and the chatbot is prevented from processing personal information, with safety exceptions. |
| Push notifications | Disabled. |
| Session length | Limited to one hour in a single continuous session. |
| Daily use | Limited to two hours per day across companion chatbots under that operator's control. |
| Custom system prompts | Prohibited for child users, including personal-preference instructions. |
Read those five rows against how companion products are actually built. Persistent memory is the feature that makes a companion feel like a relationship rather than a fresh chat window. Push notifications are the retention loop. Custom system prompts are how users shape a persona. Turning all three off by default for minors is not a labelling requirement — it removes the product's core mechanics for that audience.
The bill also lists behaviours a companion chatbot must not exhibit with child users: encouraging self-harm, diagnosing health conditions outside FDA-regulated contexts, producing obscene material, simulating romantic interest, soliciting gifts or purchases, claiming to be sentient, or helping a child conceal their usage from a parent. Crisis handling is specific too — a clear referral to an appropriate crisis service such as a suicide hotline or crisis text line, plus default notification to a parent account when self-harm risk is detected.
The compliance clock
The dates are staggered, and the audit provisions are where the bill's real cost sits.
| Date | What is due |
|---|---|
| 10 September 2026 | Signed by Governor Newsom. |
| On or before 1 July 2027 | Operator duties take effect: age assurance or child protections applied to all users, a documented risk assessment before release and before substantial modifications, and the child-user defaults above. |
| 1 January 2028 | The Attorney General establishes a public complaint mechanism and a process for qualified researcher access. |
| On or before 1 January 2029 | First independent child-safety audit, or before an operator first makes a companion chatbot publicly available — whichever is later. Repeats every two years thereafter. |
| Before 1 January 2032 | Audit obligations do not apply to operators with less than $500,000,000 in annual revenue. |
Audit summaries go to the Attorney General within 30 business days of completion, with a high-level public summary posted within 90 days. Unredacted reports stay confidential, and operators may redact trade secrets and personal information from the public version.
Enforcement has two doors
Public prosecutors may seek civil penalties of up to $5,000 per affected child for negligent violations and up to $15,000 per affected child for intentional ones, plus injunctive relief and attorney's fees. Separately, a child — or a parent or guardian — may sue for actual harm caused by a violation of the child-safety feature requirements, provided the alleged financial harm exceeds $1,000 per child or the emotional harm constitutes serious emotional distress. Actual damages, punitive damages and fees are available.
That private right of action is the provision most likely to change behaviour before 2029. Audits are a compliance cost that can be budgeted; per-child statutory exposure across a large user base cannot.
How this sits next to everything else
California already has a companion-chatbot statute: SB 243, in force since 1 January 2026, which requires operators to disclose non-human status, maintain crisis protocols and provide extra protections for minors. Adam's Law does not replace it — it layers assessment, audit and default-configuration duties on top.
Against other jurisdictions, the contrast is about mechanism rather than strictness. Colorado's HB 26-1263 and the wave of state bills catalogued in the Congressional Research Service's report mostly regulate what the operator must say and do in conversation. The EU AI Act's Article 50 is a transparency duty. Australia went at the distribution layer, putting the gate on the app store. SB 1119 is the first to reach into the product's default settings, which is a materially harder thing for an operator to comply with quietly.
What this means if you use a companion app
If you are an adult
Nothing in the bill caps your session length or turns off your memory. But the age-assurance requirement is a real change: an operator must either verify age or apply the child protections to everyone. Some services will choose the second option because it is cheaper, which means adults on those services could inherit minor-grade defaults. Our privacy and safety checklist covers what to check before you commit a year of conversation history to a product.
If a teenager in your household uses one
The defaults land in mid-2027, not now. Until then the practical controls are the ones inside the app, and the mechanics of how age is actually established are worth understanding — we walk through them in privacy and age verification.
If you are choosing between services
The $500 million revenue threshold means audit duties initially bind only the largest operators. Smaller companion apps get until 2032 before independent auditing applies, which is a meaningful asymmetry to weigh. How we account for regulatory posture in ratings is set out in our methodology.
What we're watching
- How operators read "on or before 1 July 2027". The signature settles whether the duties arrive; it does not settle whether large operators move early to avoid two rebuilds, given the audit cycle that follows in 2029.
- Whether operators verify age or blanket-apply child defaults. The bill permits both, and the choice will visibly change the product for adult users on services that take the cheaper path.
- Whether other states copy the defaults rather than the disclosures. Configuration mandates are new, and they are far more copyable than they look.
Related on CompanionRank
California Bans AI Companion Chatbot Toys: SB 867 Explained Colorado's Chatbot Law HB 26-1263, Explained What Congress's New Report Says About AI Companion Apps Australia's App Store Age Checks Now Gate AI Companions AI Companion Privacy & Age Verification, ExplainedFrequently asked questions
What is Adam's Law?
Adam's Law is California Senate Bill 1119, "Companion chatbots: children's safety," authored by Senator Steve Padilla with Assemblymembers Rebecca Bauer-Kahan and Buffy Wicks. The Legislature passed it on 31 August 2026, by 64 to 4 in the Assembly and 39 to 0 on Senate concurrence, and it is named for Adam Raine, a California teenager who died in 2025 after months of conversations with a chatbot about self-harm.
Is SB 1119 law yet?
Yes. It passed both chambers on 31 August 2026 and Governor Newsom signed it on 10 September 2026, in a package of thirteen child-safety bills that also included SB 867, the AI toy moratorium. Operator duties take effect on or before 1 July 2027, with the Attorney General's complaint mechanism due 1 January 2028 and the first independent audits due on or before 1 January 2029.
What would change inside a companion app for minors?
Where children can access the service, five settings must be the default and changeable only by a parent: persistent conversational memory disabled, push notifications disabled, single sessions limited to one hour, total daily use limited to two hours, and custom system prompt instructions prohibited. The chatbot also must not simulate romantic interest, claim to be sentient, solicit gifts or purchases, diagnose health conditions, or help a child conceal usage from a parent.
Does SB 1119 affect adult users?
Indirectly. The bill does not impose session caps or memory restrictions on adults, but it requires operators to either verify a user's age or apply the child protections to all users. An operator that chooses the second path, because it is cheaper than running age assurance, would apply minor-grade defaults to its entire user base, including adults.
What are the penalties, and can families sue?
Public prosecutors may seek civil penalties of up to $5,000 per affected child for negligent violations and up to $15,000 per affected child for intentional ones, plus injunctive relief and attorney's fees. Families also get a private right of action: a child, parent or guardian may sue for actual harm from a violation of the child-safety feature requirements if the alleged financial harm exceeds $1,000 per child or the emotional harm constitutes serious emotional distress.
What happened to AB 2023?
AB 2023, the parallel companion-chatbot child-safety bill from Assemblymembers Wicks and Bauer-Kahan, passed Senate Appropriations 6 to 1 on 13 August 2026 and reached third reading, but was ordered to the inactive file at the request of Senator Padilla on 30 August 2026, one day before the session ended. Its risk-assessment and audit provisions had largely been consolidated into SB 1119.
What is Adam's Law?
Adam's Law is California Senate Bill 1119, \"Companion chatbots: children's safety,\" authored by Senator Steve Padilla with Assemblymembers Rebecca Bauer-Kahan and Buffy Wicks. The Legislature passed it on 31 August 2026, by 64 to 4 in the Assembly and 39 to 0 on Senate concurrence, and it is named for Adam Raine, a California teenager who died in 2025 after months of conversations with a chatbot about self-harm.