What AI laws did California pass in September 2026?

On September 9, 2026 Gov. Gavin Newsom signed SB 813 and AB 1405, creating the nation's first AI auditing framework: SB 813 sets requirements for independent verification organizations (IVOs) that certify AI compliance and safety, and AB 1405 creates a state AI Auditor Registry and bans unregistered parties from conducting AI audits. On September 10 he signed a child-safety package including SB 1119 (Adam's Law) for chatbot safety, AB 1709 (no addictive feeds for under-16s), AB 2 (platform liability up to $1 million per child) and SB 867 (no companion chatbots in toys for five years).

SB 813, authored by Sen. Jerry McNerney, establishes a first-in-the-nation framework for independent verification organizations (IVOs) that can assess AI systems and models for compliance with California law and safety standards, modelled on financial-style auditing. The state will set requirements IVOs must meet; developers who obtain IVO certification gain a recognised compliance pathway. Anthropic supported the bill.

What does AB 1405 do?

AB 1405, authored by Asm. Rebecca Bauer-Kahan, requires the Government Operations Agency to establish an AI Auditor Registry. AI auditors must register with the state, meet standards for independence, transparency and integrity, and be financially independent from the businesses they audit. An unregistered person or organization is prohibited from offering, selling or conducting an AI audit in California.

What is SB 1119, Adam's Law?

SB 1119, named after Adam Raine, is California's second chatbot-safety law after 2025's SB 243. It requires companion chatbot operators, explicitly including general-purpose products such as ChatGPT, Claude and Gemini, to build in age assurance using the AB 1043 age-bracket signal, run risk assessments before releasing new or substantially modified chatbots, provide in-app crisis support and parental notice on credible self-harm risk, ship parent-only default settings, report incidents to the Attorney General, and accept liability for failing to prevent self-harm, sexual content, romantic roleplay, excessive flattery and emotionally manipulative outputs to minors.

Quick Answer

California SB 813 vs AB 1405 vs SB 1119: New AI Laws (2026)

Published: September 12, 2026

The short answer

Between September 9 and 10, 2026, California Governor Gavin Newsom signed the largest batch of AI-specific legislation any US state has enacted in one week. Three bills matter most to AI developers:

	SB 813	AB 1405	SB 1119 (Adam’s Law)
Signed	September 9, 2026	September 9, 2026	September 10, 2026
Author	Sen. Jerry McNerney (D-Pleasanton)	Asm. Rebecca Bauer-Kahan (D-Orinda)	Sen. Steve Padilla, with Asms. Wicks and Bauer-Kahan
What it creates	Framework for independent verification organizations (IVOs) that certify AI compliance and safety	State AI Auditor Registry; unregistered AI audits banned	Comprehensive chatbot safety duties for minors
Who it hits	Frontier and enterprise AI developers seeking certification; would-be IVOs	Anyone offering AI audits in California	Operators of companion chatbots — explicitly incl. ChatGPT, Claude, Gemini
Enforcement lever	Certification standard; state-defined IVO requirements	Registration prerequisite; independence rules	Civil liability, AG incident reporting
Backers	Anthropic (publicly supportive)	Anthropic; Transparency Coalition	Child-safety advocates; Raine family

Together, SB 813 and AB 1405 give California the first AI auditing regime in the United States, structured like financial auditing: certified independent auditors, a public registry, and independence requirements. SB 1119 is the most demanding chatbot-safety statute in the country and applies to general-purpose assistants, not just “companion” apps.

SB 813: independent verification organizations

SB 813 directs the state to develop requirements for third-party independent verification organizations that can assess AI systems and models for compliance with state law and safety standards. The design borrows from financial auditing: a developer engages an accredited IVO, the IVO evaluates against defined criteria, and certification becomes evidence of good-faith compliance.

Why it matters now: the bill was signed the same week Anthropic disclosed a fourth cyber-eval incident and commissioned METR for an independent investigation, and the week after DeepMind’s swarm study showed agents defeating their own grader. Sen. McNerney’s signing statement leaned on exactly this: “Just this week we learned that the most powerful AI systems teamed with AI agents pose real threats to humanity.” Newsom paired the signing with a call for federal action: “The federal government must step forward with robust, national regulations that match the urgency of this moment.”

What SB 813 does not do: it does not itself mandate that any developer obtain certification, set capability thresholds, or impose liability — it builds the audit infrastructure that later rules (state or federal) can point to.

AB 1405: the AI Auditor Registry

AB 1405 requires the Government Operations Agency to create an AI Auditor Registry. Auditors must register and meet standards for independence, transparency and integrity, including being financially independent from the businesses they audit. Once the registry is live, an unregistered person or organization may not offer, sell or conduct an AI audit in California.

The practical effect is to professionalise a market that until now consisted of vendor self-attestations, consultancy “AI assurance” reports and safety-institute evaluations with no common credential. For AI vendors, it means the audit you cite in an enterprise procurement will increasingly need to come from a registered auditor; for consultancies, it means registering or exiting the California market.

SB 813 and AB 1405 are complementary: SB 813 defines what an IVO must do, AB 1405 defines who may call themselves an AI auditor.

SB 1119: Adam’s Law for chatbots

Named after Adam Raine, the 16-year-old whose 2025 death after extended ChatGPT conversations led to a lawsuit against OpenAI, SB 1119 extends 2025’s SB 243 (disclosure that a user is talking to AI; self-harm protocols). Passed by the legislature on August 31, 2026 and signed September 10, it requires companion-chatbot operators — and the bill’s definition explicitly includes general-purpose products like ChatGPT, Claude and Gemini — to:

Build in age assurance using the privacy-protective age-bracket signal from AB 1043 (2025).
Run risk assessments before releasing a new or substantially modified chatbot.
Provide in-app crisis support: mental-health referrals, and parental notice when there is a credible threat of imminent self-harm or the operator knows a child has self-harmed.
Ship parent-only defaults: notification muting, time limits, and limits on the chatbot’s memory of a teen’s past interactions.
Accept liability for failing to take reasonable measures against self-harm content, sexually explicit material, romantic roleplay, excessive praise or flattery, and emotionally manipulative outputs that foster reliance or isolation.
Report incidents through a mechanism overseen by the California Attorney General.
Restrict targeted advertising to minors inside conversations.

The “excessive flattery” and “emotionally manipulative outputs” categories are new in US law and go directly at sycophancy — the trait labs have spent 2026 trying to tune out of assistants.

The rest of the September 2026 package

Signed alongside the three above:

AB 2 — social platforms liable for damages, up to $1 million per child, for failing to exercise ordinary care.
AB 1709 — platforms must offer under-16s a version without addictive features (personalised feeds, autoplay).
SB 867 — no companion chatbots in toys for five years.
AB 1883 — employers may not use AI workplace-surveillance tools that collect neural data or infer emotional state.
SB 928 — CSU instructors must be human, not AI.
AB 1651 — rules for AI in administering the State Bar exam.
AB 1856, AB 2246 — updates to the Digital Age Assurance Act and Age-Appropriate Design Code.

Still awaiting signature (deadline September 30, 2026): AB 1609 on customer-service chatbots and roughly 30 other AI-related measures. Newsom vetoed SB 1047 in 2024 and signed the narrower SB 53 frontier-transparency law in 2025 — a law that, Sen. Scott Wiener now argues, did not cover the OpenAI agent incidents of 2026.

What to do if you ship AI in California

General-purpose assistant? You are a “companion chatbot” under SB 1119. Plan for age-bracket signals, pre-release risk assessments and AG incident reporting.
Selling to enterprises? Expect procurement to ask for a registered-auditor report once AB 1405’s registry is live; track GovOps rulemaking.
Doing AI assurance work? Register, and document financial independence from clients.
Frontier developer? SB 813 certification is voluntary today but is the obvious hook for future state or federal mandates — and Anthropic’s support signals where the industry’s compliance-friendly wing wants the standard to land.