Everyone cites the wrong California law for AI calls

California AB 2905 has applied since 1 January 2025. It requires telling the person called that a prerecorded message uses an artificial voice. It applies to calls placed through an automatic dialing-announcing device. California's Public Utilities Code defines that as equipment that stores or generates numbers to be called.

Read those two phrases again: numbers to be called, calls placed. Both describe a business dialing out. A receptionist that picks up when your customer rings you isn't dialing anyone, so AB 2905 does not govern it.

Several vendor pages still apply AB 2905 to inbound calls, some with a $500-per-call figure attached. If you've read one of them, you've read a claim the statute text doesn't support.

SB 1001, known as the B.O.T. Act, has been operative since 1 July 2019. It bans using a bot online to push a sale or a vote while hiding that it's a bot. Disclosure is the cure.

Look at the word online. SB 1001 defines a bot as an automated online account, and online as appearing on a public website, web app or digital app. A phone call to your business line is none of those.

SB 1001 still matters if you run a chat bot on your website, because that bot is an online account. For the calls your business answers, look to Maine and Utah.

Maine's Title 10 §1500-DD has applied since 1 October 2025. A business may not let an AI chatbot mislead a reasonable consumer into thinking they're dealing with a human. Clear and conspicuous notice fixes that. Maine's definition covers aural conversation, so it reaches an AI receptionist on the phone.

Utah's AI Policy Act, SB 149 as amended by SB 226, has applied since 7 May 2025. Most businesses must disclose when a consumer asks. State-licensed occupations such as medicine, law, accounting, nursing and psychology must disclose prominently and verbally at the start of the conversation.

GreetKeeper handles both with one setting. The caller hears, in your own wording, that they're speaking with an AI receptionist before anything else happens. A dental practice in Salt Lake City and a plumber in Portland, Maine, use the same setting. Every caller hears it on the first call.

In February 2024 the FCC ruled that an AI-generated voice counts as an artificial voice for calls a business places to consumers. Damages under that framework run from $500 to $1,500 per violation. That's why it gets quoted so often, and why it gets pinned on inbound products it was never written for.

In Insurance Marketing Coalition v. FCC, the FCC one-to-one consent rule was struck down by the Eleventh Circuit on 24 January 2025. It was later repealed. Pages still describe it as live.

A September 2024 FCC proposal on AI disclosure and a two-second opt-out has not become law. Calling it a requirement is the same mistake as the California one: a real document, cited for something it doesn't do.

This is how we read the statute texts in September 2026. Bring it to your counsel, who has the final word, and ask which state rules reach the callers you actually get.

Texas and Colorado both have measures in motion. Ask counsel for the current text before you rely on any summary, including ours. Keep the disclosure setting on in the meantime; it costs you nothing on a call.