Governor Gavin Newsom signed AB 1609 on 28 September. It becomes Chapter 733 of the Statutes of 2026 and adds a new chapter to California’s Business and Professions Code, and it does something narrower and more enforceable than most AI legislation: it sets a service standard for customer support chatbots and attaches a fine to missing it.
The law applies to businesses with more than $500,000,000 in gross annual revenue nationally that provide goods and services to customers. That threshold excludes almost every startup and captures the retailers, carriers, banks and airlines where automated support is already the default.
Two duties
The first is disclosure. A covered business must not represent that a customer service chatbot is a human, and must give “a clear and conspicuous disclosure that the customer service chatbot is artificially generated and not human” where a reasonable person might otherwise be misled.
The second is escalation, and it is the part with numbers in it. During regular business hours a covered business must offer a simple way for a customer to ask to be connected to a human agent. It must then make good faith efforts either to connect the customer within 15 minutes of the request, or to schedule a specific appointment time within one business day. Across telephone and online platforms combined, cumulative hold time may not exceed one hour in total.
Enforcement sits with public prosecutors. Civil penalties are capped at $5,000 for an initial violation and $10,000 for each subsequent violation. The statute creates no private right of action, so customers cannot sue over it themselves. Compliance dates in the bill are set from 1 January 2027.
Why the escalation clause matters more than the label
Disclosure rules for chatbots are not new; California already has one for bots in certain commercial and political contexts, and the EU AI Act requires that people be told when they are interacting with a machine. What AB 1609 adds is a right to leave the machine, on a clock.
That converts a transparency requirement into an operational one. A company can satisfy a disclosure rule with a line of small print. It cannot satisfy a 15-minute handoff target without staffing a queue. The cost of automated support, for firms over the revenue line, now includes the cost of the humans the automation was meant to replace.
Part of a larger package
AB 1609 is one of several AI measures California enacted this month. On 10 September the governor signed SB 1119, which sets a safety framework for companion chatbots with risk assessments and independent audits from 2029, and SB 867, a moratorium through 2031 on companion chatbots in toys. On 9 September he signed SB 813, requiring a state agency to set up a process for selecting and regulating independent AI verification organisations by 1 January 2028, and AB 1405, mandating an AI auditor registry by 1 January 2029.
The policy direction is consistent: California is not trying to license models, it is regulating the points where automated systems touch the public — toys, companions, customer service desks — and building the audit infrastructure to check the claims made about them.
What to watch
The governor’s deadline for the rest of the session’s bills is 30 September, and several AI measures are still awaiting action, including SB 947 on automated decisions in employment and SB 1000 on AI transparency. The other thing to watch is how the 15-minute standard is measured in practice: the statute requires good faith efforts rather than a guaranteed connection, and the first enforcement action will decide what that phrase is worth.
