California AI Laws Explained: SB 53 and the Rest

Short answer: California has no single AI law. It has many separate laws and rules, and each covers a different group. Senate Bill (SB) 53, the Transparency in Frontier Artificial Intelligence Act, has been in force since January 1, 2026. It reaches only the builders of the very largest AI models.

Most ordinary businesses are more likely to meet the chatbot, ad, privacy-agency and hiring rules in the table below. Several of those start in 2027. This is not legal advice. Check the current text before you rely on it.

TL;DR

• Status on October 5, 2026: SB 53 is in force. The code sections show no amendment. The Legislature's pages, the code site and the Center for AI and Digital Policy's case page showed no order stopping it that day.
• SB 53 covers a developer whose model took more than 10^26 (a 1 followed by 26 zeros) basic computer arithmetic steps to train. Only the state attorney general (AG) can recover its penalty, up to $1,000,000 per violation, and it falls on developers with over $500 million in yearly revenue.
• An ordinary business is more likely to meet the bot, ad-disclosure, automated-decision, hiring and chatbot rules. Which ones depends on what you build or use.
• The law moved fast this fall. SB 1119 (September 10), SB 1050 (September 16), and SB 947, SB 1000 and AB 2713 (September 30) are all new.
• Check: each law's bill page and the code site before you act. Links are at the end.

Why there is no single California AI law

A statute is a law passed by the Legislature. A code is the state's organized set of statutes. A regulation is a binding rule an agency writes under a statute. California's AI rules come in all three forms. For how a signed bill lands in the code, see what a session law is. For statutes versus rules, see what a regulation is.

That is why a search for "the California AI law" returns a mix. Some pages mean SB 53. Some mean the privacy agency's rules. Some mean a chatbot law. They have different readers.

The laws, in one table

I checked each law on October 5, 2026 against the signed bill, the current code and the agency rule page. A "chaptered" bill is the final version the governor signed and the state numbered. This list is not every California law that touches AI. Others cover health care, elections, digital likenesses, schools and AI auditors.

How to read the names. SB means Senate Bill and AB means Assembly Bill. "Bus. & Prof. Code § 22757.10" means section 22757.10 of the Business and Professions Code. "Civ. Code" is the Civil Code and "Lab. Code" is the Labor Code. "Cal. Code Regs. tit. 11" means title 11 of the California Code of Regulations, the state's collection of agency rules. "Generative AI" means AI that makes text, images, audio or video.

One note on dates. SB 1000 is an urgency statute, meaning it takes effect the day it is signed. The other September 2026 bills I read have no urgency clause. Article IV, section 8(c) of the California Constitution says an ordinary statute "shall go into effect on January 1 next following a 90-day period from the date of enactment." That gives January 1, 2027, unless the bill sets another date. SB 1050 has no date of its own, so I used that rule.

Which ones reach an ordinary business

Start with what your business does.

• You use a chatbot to answer customers. Section 22601 excludes bots "used only for customer service." But § 17941 still bars hiding a bot you use to push a sale.
• You hire in California with a screening tool. The Civil Rights Council rules apply now. SB 947 adds duties in July 2027. If you also fall under the California Consumer Privacy Act, the privacy agency's rules apply from January 1, 2027.
• You run video or audio ads. SB 1050 applies if an AI-made performer is prominent.
• You build or retrain AI for the public. AB 2013 and the AI Transparency Act may apply. A chatbot built to keep people company brings SB 243.
• You only use a big vendor's model. SB 53 does not reach you. The vendor carries those duties.

Here is one worked case. A 200-person California employer screens resumes with a tool, runs a support chatbot, and posts a video ad with an AI actor. The Civil Rights Council rules cover the resume tool today. SB 947 covers tools used for discipline or firing, so it would not reach resume screening. The support bot is outside SB 243 if it only handles service, but § 17941 applies if it pushes sales and hides what it is. The ad needs SB 1050's note from 2027. SB 53 never comes up.

AB 316 applies to everyone. In a suit over harm caused by AI, the person who built, changed or used it "may not assert" that "the artificial intelligence autonomously caused the harm." That is Civil Code § 1714.46, in force since January 1, 2026.

SB 53 from the statute

SB 53 is Chapter 138 of the Statutes of 2025. The governor approved it on September 29, 2025. It adds Business and Professions Code sections 22757.10 to 22757.16.

Who it covers. Section 22757.11(i) says a "frontier model" is "a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations" (counts of basic arithmetic steps). A foundation model is a general-purpose AI trained on broad data. Section 22757.11(j) says a "large frontier developer" had revenue, together with its affiliates (related companies under common control), above "five hundred million dollars ($500,000,000) in the preceding calendar year."

What it requires. Large frontier developers must write, follow and publish a safety framework, and review it every year (§ 22757.12). Every frontier developer must post a transparency report when it releases a new model. Every frontier developer must report a "critical safety incident" to the state Office of Emergency Services "within 15 days of discovering" it (§ 22757.13(c)). That means serious harm, such as a model slipping out of its maker's control and causing death or injury. If an incident poses an imminent risk of death or serious injury, the developer must tell the right authority within 24 hours.

Penalty and enforcement. Section 22757.15(a) sets a civil penalty, a fine paid to the state, of up to $1,000,000 per violation. It names a "large frontier developer." Section 22757.15(b) adds: "A civil penalty described in this section shall be recovered in a civil action brought only by the Attorney General." A civil action is a lawsuit. I found no section in the chapter that lets a private person or company sue.

Whistleblowers. Labor Code § 1107.1 bars a frontier developer from blocking or punishing a "covered employee" (one who works on assessing or managing safety risk) who reports a catastrophic risk or a violation. That employee can file a lawsuit, and the court may order the developer to pay the employee's lawyer fees.

Rules. The text lets the Office of Emergency Services name federal standards by regulation (§ 22757.13(h)). I found no other power in the chapter to write binding rules.

The dates, in order

• 2018 to 2024: SB 1001 (the bot rule), SB 942 and AB 2013 were enacted. AB 2013 took effect January 1, 2025.
• September 29, 2025: SB 53 approved.
• October 13, 2025: SB 243 (chapter 677) and AB 853 (chapter 674) signed. AB 853 pushed the AI Transparency Act's start to August 2, 2026.
• January 1, 2026: SB 53, SB 243, AB 316 and the privacy agency's rules took effect. AB 2013 postings were due.
• March 4, 2026: a federal judge refused to block AB 2013 in X.AI v. Bonta, says the Center for AI and Digital Policy. It also reports an appeal in the Ninth Circuit, the federal appeals court for California (No. 26-1591), with argument set for November 18, 2026.
• August 2, 2026: the AI Transparency Act took effect.
• September 10, 2026: SB 1119 (chapter 190) adds a child-safety chapter for companion chatbots.
• September 16, 2026: SB 1050 (chapter 246) signed.
• September 30, 2026: SB 947 (chapter 859), AB 2713 (chapter 856) and SB 1000 (chapter 861) signed. AB 2713 revises the duties of large online platforms. SB 1000 removes the old 1,000,000-user threshold for companies covered by the AI Transparency Act and renames the free tool a "disclosure verification tool."
• January 1, 2027: state reports on SB 53 begin. The privacy agency's automated-decision duties start. SB 1050 and the other ordinary September bills start.
• July 1, 2027: SB 947 operates, and most of SB 1119's duties start.

On the federal side, a December 11, 2025 executive order (90 Federal Register 58499) tells the Justice Department to challenge state AI laws. An executive order is not a court ruling and does not change the California code.

What a business should check

1. List every chatbot, hiring tool, scoring tool and AI ad you use or build.
2. Match each to the table. Note whether you are the builder or just the buyer.
3. Ask each vendor which of these laws it says it meets. Ask for the document.
4. For hiring and lending tools, find out if a human reviews the output and can change it. The privacy agency's rules turn on that.
5. Put January 1 and July 1, 2027 on your calendar.
6. Open the bill pages and the code site every quarter. In three weeks this fall, two of these laws changed and new ones arrived.

What goes wrong

The common mistake is calling SB 53 "the California AI law" and stopping there. It does not touch most firms. The next is old numbers. Some write-ups give a "$100 million" revenue line for SB 53. The enacted text says $500 million. Write-ups from before September 30, 2026 say the AI Transparency Act needs over 1,000,000 users. SB 1000 removed that line. The last is reading the privacy agency's January 1, 2026 date as the start of the automated-decision duties. Section 7200(b) says a business "must be in compliance with the requirements of this Article no later than January 1, 2027."

For the sister law in New York, see the New York RAISE Act explained. For another state's approach, read the Colorado AI law guide. To cite a California section correctly, see how to find and cite California law.

Where to read the official text

The signed SB 53 is on the Legislature's bill page. Every other bill here has a page of the same form with its number in place of "SB53." The California codes site shows the current text of each section, though it can trail a new bill by days. The privacy agency's rules page lists adopted rules and showed no proposed ones on October 5, 2026.

This guide is part of US Law Data: The Complete Guide, a map of where US law comes from and how to use it.

FAQ

Is there a California AI law in effect? Yes, several. SB 53, SB 243, AB 316 and the privacy agency's rules took effect on January 1, 2026. The AI Transparency Act took effect on August 2, 2026. No one law covers all AI.

Does SB 53 apply to my small business? Almost certainly not. It reaches developers who trained a model with more than 10^26 operations. If you only use someone else's model, those duties sit with the developer.

When did SB 53 take effect? January 1, 2026. The code sections show "Added by Stats. 2025, Ch. 138" and no later amendment.

Can I sue under SB 53? The penalty can be recovered only by the attorney general (§ 22757.15(b)), and the chapter has no section letting private parties sue. The separate whistleblower law is different. An employee who is punished for reporting a risk can file a lawsuit under Labor Code § 1107.1.

Do I have to tell customers they are talking to AI? It depends on what the bot does. If it is used to sell, § 17941 bars hiding that it is a bot. If it is a companion chatbot, SB 243 requires a notice when a person could mistake it for human. A pure customer-service bot is outside SB 243.

Do the privacy agency's automated-decision rules apply to me? Only if you are a business under the California Consumer Privacy Act and use the technology to make a significant decision. The rules define that as a decision on lending, housing, school, jobs or contract work, or health care. The duties begin January 1, 2027.

Has the federal government blocked California's AI laws? An executive order from December 2025 tells the Justice Department to challenge them. On October 5, 2026, the Legislature's pages, the code site and the Center for AI and Digital Policy's case page showed no court order stopping SB 53. Check the court dockets if it matters to you.

If your product needs current US statute and regulation text, with its amendment history, see the US primary law API, a data feed that software can read.

Priyansh Khodiyar

Co-Founder & CTO

Priyansh leads engineering and AI at Vaquill AI: the pipelines that pull statutes, regulations and court rules from every US jurisdiction's official publisher, and the REST API, MCP server and open dataset that serve them.