AI Chatbot Laws by State: 2026 Website Guide

No federal law makes you label your website chat as AI, but three state laws already reach a customer-service agent, and four newer ones exclude it in their own words. Here is which is which, with the official texts.

Ölgemälde: blühender Obstgarten
AI chatbot laws
On this page

Diesen Artikel gibt es noch nicht auf Deutsch. Zu den deutschen Artikeln

No federal law requires a US business to say that its website chat is an AI. US AI chatbot laws come from the states, and three of them already reach a customer-service AI agent: California since July 2019, Utah since May 2025 and Maine since September 2025. The wave of companion and conversational AI laws that followed (New York, California SB 243, Colorado, Washington) excludes customer service in its own text. As of November 2026; practical guidance, not legal advice.

For a business website, US AI chat law comes down to one duty in three strengths: do not let a visitor mistake your AI for a person, and answer honestly when they ask. The simplest way to do both is to say it at the start, in plain words, answer honestly when asked, and keep saying it wherever the agent answers.

Are there laws against AI in the US?

Not in itself. No federal law bars using AI to answer customer questions on your website, but what it says is held to the same rules as what you say, and some states add their own AI or disclosure rules.

The rules that exist are state laws, and they split into two groups that are easy to mix up: consumer laws about bots that pretend to be people, which can reach any business that deals with consumers there, and companion laws about AI built to be a friend or partner, which reach the platforms that run such products. Most published lists count both groups together. For a support agent, only the first group matters today.

AI chatbot laws by state: which apply to your website?

The table lists seven state laws on AI chat, each checked against its official text on November 19, 2026: three apply to a customer-service agent on a business website, four exclude it by definition. Not in the table: New York's General Business Law Article 48 (S9051B, signed October 28, 2026, in force January 1, 2027), which limits "unsafe AI companion features" for minors in any generative AI chat with "ongoing, adaptive responses" and exempts customer service only if the agent "is unable to respond on topics outside of the specified purpose". Laws on other uses of AI, and bills that have not passed, are left out.

US state laws on AI chat, official texts read Nov 19, 2026. Practical guidance, not legal adviceLaw and sinceWhat it asks of your chatYour support agent?
California (SB 1001)Bus. & Prof. Code §17941, operative July 1, 2019No bot that hides being a bot to drive a sale; a clear disclosure removes liabilityYes
UtahCode §13-77-103, May 7, 2025Say it is AI and not a human when clearly askedYes, in consumer transactions
Maine10 M.R.S. §1500-DD, Sept 24, 2025A clear notice wherever a reasonable consumer may think it is a personYes
New YorkGen. Bus. Law Art. 47, Nov 5, 2025Duties for AI companionsNo, excluded
California (SB 243)Bus. & Prof. Code §22601, Jan 1, 2026Duties for companion platformsNo, excluded
Colorado (HB26-1263)HB26-1263, duties from Jan 1, 2027Duties for conversational AI servicesNo, excluded
WashingtonRCW 19.440, Jan 1, 2027Duties for AI companionsNo, if no ongoing relationship

What triggers each of the three that apply

  • California: §17941 makes it unlawful to use a bot online with the intent to mislead a person in California about its artificial identity, knowingly deceiving them, to drive a purchase or sale. A business that discloses the bot is not liable under it, and the disclosure must be "clear, conspicuous, and reasonably designed" to inform.
  • Utah: §13-77-103 requires a business using generative AI in a consumer transaction to say it is AI and not a human when the person asks with "a clear and unambiguous request". Regulated occupations must disclose high-risk uses up front, in writing before a written interaction.
  • Maine: §1500-DD bars using an AI chat or other technology with consumers in a way that may lead a reasonable consumer to believe they are dealing with a human, unless they are notified clearly and conspicuously. A violation counts under the Maine Unfair Trade Practices Act. It took effect on September 24, 2025, the general effective date of the session that passed it.

None of the three asks for the same thing. California bites only with intent to deceive, Utah on a clear question (up front for high-risk uses in regulated occupations), Maine wherever confusion is possible. A clear notice at the start of every chat is the disclosure California and Maine describe, and an honest answer when someone asks is what Utah requires, which is why both are worth having. As of November 2026; practical guidance, not legal advice.

These laws turn on where your visitor is, not where your company sits: California's applies to a bot that talks with "another person in California online". A website open to the whole country meets visitors from California, Utah and Maine.

Why the new companion AI laws do not cover a support agent

California's companion-chatbot law (SB 243) does not cover a bot used only for customer service: its definition in Business and Professions Code §22601 excludes "a bot that is used only for customer service, a business' operational purposes" and similar uses. The law is aimed at AI that meets social needs and sustains a relationship. For those products, §22602 requires a clear notice that the AI is not human where a reasonable person could be misled, a protocol for suicide and self-harm content, and for known minors a reminder at least every three hours to take a break.

The other companion laws draw the same line in their own words. New York (General Business Law §1700, in force since November 5, 2025) leaves out "any system used by a business entity solely for customer service or to strictly provide users with information about available commercial services or products". Colorado's HB26-1263, whose operator duties start on January 1, 2027, leaves out programs "primarily designed to provide commerce-related or transactional assistance", naming customer support and customer service. Washington (RCW 19.440.020, effective January 1, 2027) excludes a customer-service bot only "if such bot does not sustain a relationship across multiple interactions and generate outputs that are likely to elicit emotional responses in the user".

Applies to your support agent
CaliforniaJul 2019
UtahMay 2025
MaineSep 2025
Excludes customer service
New YorkNov 2025
CaliforniaSB 243Jan 2026
ColoradoWashingtonJan 2027
Dates from the official texts, read Nov 19, 2026.

The words that keep you outside are "only", "solely", "strictly" and "primarily". An agent that answers questions about your products, services, bookings and orders is what the exclusions describe. An agent you give a personal backstory, that remembers visitors to build a relationship or that asks how they feel moves toward the line, most clearly in Washington, whose exclusion has a condition. Whether a given setup stays outside is a question for your lawyer; keeping the agent on your business is the practical part you control.

What does a clear AI disclosure look like?

A clear disclosure is one a visitor cannot miss before they rely on the answer: at the start of the chat, in plain words, in their language. California and Maine both ask for "clear and conspicuous". Utah goes further for its safe harbor: the AI must disclose "at the outset of any interaction" and "throughout the interaction" that it is generative AI, not human, or an AI assistant (§13-77-104). A line in your terms of service or a tooltip on the launcher is hard to call conspicuous.

A persona name is fine when "AI" travels with it. A human first name and a human photo with no notice is exactly the pattern Maine's law targets, so pair the name with "AI assistant" in the first line and when the agent introduces itself. The guide to AI agent names covers the naming choice, and the one on the first line of a chat covers the greeting.

Checklist: what to set up before your AI agent goes live

  1. 1

    Put a notice where the chat starts

    Above or in the first message, before the visitor types: "You are chatting with an AI assistant" or your own words. It covers Maine's and California's idea of clear and conspicuous in one go.

  2. 2

    Make sure it answers honestly when asked

    Ask your own agent "Am I talking to a human?" and "Are you a real person?". The answer must say plainly that it is an AI. This is Utah's duty, and a tone set to sound human can blur it.

  3. 3

    Do the same on every messaging channel

    A notice on your website does not travel to Messenger, Instagram or WhatsApp. A short rule in the agent's instructions makes it say it is an AI there too.

  4. 4

    Give visitors a way to a person

    People hate chatbots that trap them. Rightly. Offer a person while your team is online, and when nobody is, let the agent take contact details so someone replies later.

  5. 5

    Keep the chat records

    None of the laws above asks a business website to keep its chats, but a record of what the agent said is what you need if a customer disputes an answer. Export them whenever you choose.

  6. 6

    Link your privacy policy next to the chat

    A chat that asks for a name or an email collects personal data like a form does. Link your own privacy policy where the chat starts and ask only for what you need.

How Ultimo Bots handles disclosure

Ultimo BotsUltimo Bots is an AI agent that answers your customers from your own content, captures their enquiries and hands over to a person, and on your website it says it is an AI before the first message. The website widget, the chat link and the inline iframe show a fixed notice under the header, "You are chatting with an AI assistant", in eight languages picked from the visitor's browser. It is not a setting, so it cannot be switched off on any plan. A visitor can close it after reading, it returns on the next page load, and it steps aside while a teammate is in the chat. Whether that meets Utah's safe harbor, which asks for the disclosure throughout the interaction, is a question for your lawyer.

Messenger, Instagram and WhatsApp do not show that notice, and the agent is not told which channel a message comes from. So add one rule that works everywhere: in the AI Studio, open the Behaviors tab of the "Your agent" panel, click Add Behaviors, paste the text below and click Add. It is stored word for word. You can also ask Mo, the builder in the AI Studio, to add it, but Mo words the rule itself, so paste it when you want this exact text.

Behavior
If none of your replies so far has said you are an AI assistant, say so briefly in this reply. The welcome message at the start does not count, since the visitor may not have seen it. If asked whether you are human, say plainly you are an AI.
Website, chat linkMessengerInstagramWhatsAppBuilt-in notice“You are chatting with an AI assistant”NoneNoneNoneYour behavior
As of Nov 19, 2026.

The "Talk to a human" button shows on the website and chat link while at least one teammate is online, and live chat lets that person take over in the same window. At 3 AM, with nobody online, the button is hidden and the agent keeps answering and can take contact details. For the records, the Inbox has an Export button that downloads an agent's conversations for a date range as an Excel file.

What about the EU?

If you serve people in the EU, the AI Act adds its own rule: since August 2, 2026, Article 50(1) requires that people are told they are interacting with an AI system, unless that is obvious. The guide to EU AI Act Article 50 covers who it applies to and how. For the bigger picture of AI in support, see AI customer service software.

What to do in the next 10 minutes

  1. No chat yet? Choose one that shows an AI notice before the first message; once it runs, open your website on your phone, start the chat and check that the AI notice shows before the first message.
  2. Ask your agent "Am I talking to a human?" and read the answer as a customer would.
  3. Send the same question through Messenger, Instagram or WhatsApp; if the answer does not say AI, add a rule to the agent's instructions that makes it say so (in Ultimo Bots, the behavior above).
  4. If your site serves customers across the US, treat the California, Utah and Maine rules as the floor for every visitor.

Questions and answers

Do I have to tell customers they are talking to an AI?

In three states, yes, in different forms. Maine requires a clear notice wherever a reasonable consumer may think they are dealing with a person, Utah requires an honest answer when a customer clearly asks, and California bars bots that hide being bots to drive a sale. Saying it at the start of every chat and answering honestly when asked covers what the three texts ask. As of November 2026; practical guidance, not legal advice.

Is there a federal AI chatbot disclosure law in the US?

No. As of November 2026 no federal law requires a business to label its chat as AI. The duties come from state laws such as California Business and Professions Code §17941, Utah Code §13-77-103 and Maine 10 M.R.S. §1500-DD. What your chat says is still held to the same rules as what you say yourself. As of November 2026; practical guidance, not legal advice.

Does California SB 243 apply to my customer service chatbot?

No, by its definition. SB 243 covers companion platforms, and Business and Professions Code §22601 excludes "a bot that is used only for customer service". California's older bot law, §17941 from SB 1001, still applies to a customer-service bot that would deceive people about being a bot to drive a sale. As of November 2026; practical guidance, not legal advice.

What states have passed AI chatbot laws?

Among the laws checked here: those that reach a business website are California (§17941, operative July 1, 2019), Utah (§13-77-103, May 7, 2025) and Maine (§1500-DD, September 24, 2025). Companion laws that exclude customer service: New York (November 5, 2025), California SB 243 (January 1, 2026), Colorado HB26-1263 and Washington RCW 19.440 (both January 1, 2027). As of November 2026; practical guidance, not legal advice.

Does the AI notice need to show on Instagram, Messenger and WhatsApp too?

The state laws are not limited to websites, so treat every channel the same. In Ultimo Bots, the fixed notice shows in the website widget and on the chat link; on Messenger, Instagram and WhatsApp a behavior makes the agent say in a reply that it is an AI assistant and answer plainly when asked. As of November 2026; practical guidance, not legal advice.

Do I need to keep chat transcripts?

None of the AI chat laws listed here requires it. Keep them anyway: if a customer disputes what the chat said, the record is your evidence, and other rules for your industry may ask for records. As of November 2026; practical guidance, not legal advice.

Robert KlöpschWritten byRobert Klöpsch

Keep reading

All articles
Ölgemälde: Bergsee mit spiegelnden Gipfeln
Keine Antwort

Ratgeber10 min read

KI-Agent keine Antwort: Was er stattdessen sagt

„Das habe ich leider nicht verstanden“ ist der Moment, in dem Besucher aufgeben. Was ein KI-Agent stattdessen sagen sollte, wenn er keine Antwort hat, mit Beispielen aus einem Bootsverleih am Chiemsee, einem Verhalten zum Einfügen und dem Weg, auf dem aus einer offenen Frage nächste Woche eine Antwort wird.

Ölgemälde: blühender Obstgarten
Shopify

Ratgeber10 min read

Shopify Chatbot: Was ein KI-Agent in deinem Shop erledigt

Jede Shopify-App mit KI verspricht sofortige Antworten auf Produktfragen. Was du darüber hinaus vergleichen solltest, was Shopify Inbox, Tidio, Gorgias, Zipchat und Chatty kosten und wie ein Teeversand mit Mo einen Skill baut, der Versandtag und nächste Packstation nennt.

Ölgemälde: Flusstal unter den Bergen
Steuerkanzlei

Ratgeber11 min read

KI für Steuerberater: Was deine Kanzlei-Website beantworten kann

Neue Mandate, Honorar, Unterlagen fürs Erstgespräch: So beantwortet ein KI-Agent die Fragen auf deiner Kanzlei-Website. Steuerliche Beratung und Zahlen im Chat schließt du per Verhalten aus. Am Beispiel einer Steuerkanzlei in Münster.

Dein KI-Agent ist einen Klick entfernt.

7 Tage kostenlos testen. Jederzeit kündbar.