Mostly no. Every rule below is a state rule, the four states that wrote one down do not say the same thing, and one plain line at the top of your chat keeps you clear of all four.
Four states have put this in writing
California, Utah, Maine and Colorado each have language on the books that a chat widget could trip over. California is the narrowest, Maine the widest. Utah hands you a written way out. Colorado has just rewritten its own.
This is four states read on one day, not a survey of all fifty. Check your own.
This is not legal advice and it is not tax advice. Run it past your own attorney or your state attorney general’s consumer protection office. Your state licensing board may have its own rules about how a licensed trade talks to the public, starting with where your license number has to appear.
California only cares if you were trying to trick someone
Read California Business and Professions Code section 17941 on the California Legislative Information site, the Legislature’s own text. It was added by SB 1001 in 2018, took effect January 1, 2019, and became operative July 1, 2019. That page carries no update date.
The section makes it unlawful to use a bot to communicate or interact with another person in California online “with the intent to mislead the other person about its artificial identity,” for the purpose of knowingly deceiving that person about the content of the communication, in order to push a sale or sway a vote. Every one of those pieces has to be true at once. A chat that books drain jobs and never pretends to be human is not what this was written for.
Then comes the sentence that matters most. “A person using a bot shall not be liable under this section if the person discloses that it is a bot.” Subsection (b) says the disclosure has to be clear, conspicuous, and reasonably designed to tell the people it talks to that it is a bot.
One number in that chapter gets misread. Section 17940 defines an “online platform” as a site with 10,000,000 or more unique monthly United States visitors or users for a majority of months during the preceding 12 months. That term never appears in the rule itself. It sits in the definitions and in section 17942, which keeps the duty off web hosting and internet service providers. Being small does not put your site outside the chapter.
Utah makes you answer honestly when asked
Utah has the clearest instruction of the four. Its rules sit in Utah Code Title 13, Chapter 77, on generative AI consumer disclosures and enforcement, published by the Utah State Legislature. Sections 103, 104 and 105 show an effective date of May 7, 2025 and were enacted by Chapter 465 of the 2025 General Session. The definitions section was amended by Chapter 95 of the 2026 General Session and shows an effective date of May 6, 2026.
Section 13-77-103 says a supplier using generative AI with someone in a consumer transaction has to tell that person they are interacting with generative AI and not a human, if the person asks or otherwise prompts about whether AI is being used. The law adds that the question has to be a clear and unambiguous request. Utah does not make you volunteer it. It makes you stop hiding it once somebody asks straight out.
Section 13-77-104 is a safe harbor. You are not subject to an enforcement action for violating 13-77-103 if your AI clearly and conspicuously says, at the outset of the interaction and throughout it, that it is generative AI, is not human, or is an AI assistant. That is one greeting line and a widget label. The same section lets the state division write rules naming forms of disclosure that do and do not count, so watch that.
Two numbers, both in section 13-77-105. The division director may impose an administrative fine of up to $2,500 for each violation, and a person who breaks an administrative or court order issued for a violation faces a civil penalty of up to $5,000.
Utah has a stricter rule for licensed trades, in 13-77-103(2), but it only bites during a high-risk interaction. Section 13-77-101 defines that as collecting sensitive personal information such as health, financial or biometric data, giving personalized advice someone could rely on for a big personal decision, including financial, legal, medical and mental health advice, or other uses the division names by rule. Booking a water heater swap is not on that list today. Whether your licence puts you inside it is a question for your state board, not this page.
Maine drew the widest line
Maine Revised Statutes Title 10, section 1500-DD sits in Chapter 241, “Communications with Consumers via Artificial Intelligence”, published by Maine’s Office of the Revisor of Statutes. Its history line reads “PL 2025, c. 294, Section 1 (NEW); RR 2025, c. 1, Pt. A, Section 16 (RAL)”. The page notes its data was extracted on October 20, 2025.
Subsection 2 says a person may not use an AI chatbot, or any other computer technology, to engage in trade and commerce with a consumer “in a manner that may mislead or deceive a reasonable consumer into believing that the consumer is engaging with a human being” unless the consumer is notified in a clear and conspicuous manner that they are not. Subsection 3 makes a violation of that a violation of the Maine Unfair Trade Practices Act.
Maine does not ask whether you meant to mislead anybody. It asks whether the setup may mislead a reasonable person. A bot with a human name, a headshot and typing dots is that setup.
Colorado has just rewritten its own
Colorado Senate Bill 24-205 told the deployers and developers of an AI system intended to interact with consumers to make sure each one is told they are dealing with AI, unless it would be obvious to a reasonable person. Senate Bill 25B-004, signed August 28, 2025, moved that start date to June 30, 2026, and that date has now passed. Senate Bill 26-189, signed May 14, 2026, then repealed and reenacted that whole part of the statute around automated decisions rather than a chat window saying hello, effective January 1, 2027. House Bill 26-1263, signed May 29, 2026, sets rules for conversational AI from January 1, 2027 and leaves out chat built mainly for shopping, ordering, payments, returns and customer service. None of that is legal advice. In Colorado, put it on your attorney’s list before January.
Lying about it was already covered
On September 25, 2024, the Federal Trade Commission published “FTC Announces Crackdown on Deceptive AI Claims and Schemes”, announcing five law enforcement actions. Lina M. Khan, the chair at the time, said in it: “Using AI tools to trick, mislead, or defraud people is illegal,” and “The FTC’s enforcement actions make clear that there is no AI exemption from the laws on the books.” Those five actions were about AI claims and AI tools sold to deceive, not about a chat window that failed to say it was a bot.
Your own state’s consumer protection law is the other backstop. Maine wrote no fresh penalty for chatbots. It sent a violation into the unfair trade practices law the state already had. Ask your attorney what that law is called where you work.
The risk was never the bot. The risk is a bot that claims to be a person, or one that promises something you cannot deliver. That second half is its own problem, and it is covered in what happens when your chatbot makes a promise.
An honest limit: not one of these statutes gives you approved wording or an official badge. Each leaves what counts as clear and conspicuous open, so anyone selling court-proof phrasing is guessing.
What to do this week
Five plain steps, none of them costing anything:
- Put it in the first message the chat sends. Something like “Hi, this is an AI assistant for [your company name]. A person can call you back if you would rather.”
- Keep a short label on the widget itself, visible without opening a menu or a policy page, in the same spot as any note that the chat is recorded.
- If the bot has a name, keep the AI label right next to the name. A name on its own reads as a person.
- Make sure it answers honestly if somebody types “are you a real person”. Ask it yourself and watch.
- Write down the date you checked and what it said. If anyone asks, that note is your evidence.
Then test it before you turn it loose, using the steps in how to test your website AI agent. An AI assistant on a trade website is there to do one job, getting real work on the books for a real person to run. Saying out loud what it is does not get in the way of that.