AI Chatbot Disclosure Laws by State: When a Bot Must Say It Is AI

Short answer: there is no single US rule. Each state picks a trigger. Utah makes a bot admit it is AI when the user asks. Maine, Idaho, Oregon and California (for sales bots and companions) require it when a person could think they are talking to a human. New York, Washington and Georgia require a notice at the start and again every few hours. This page covers ten states, each read from the law itself as of 12 October 2026.

Written and checked on 12 October 2026. This is general information, not legal advice; check the cited statute or opinion before relying on it.

A chatbot is software that holds a conversation by text or voice. A companion chatbot is one built to keep up a human-like relationship over time, like a virtual friend. A disclosure is the notice that says "this is AI."

Official text, as of 12 October 2026

These three sentences, one per trigger, were read from the official pages on 12 October 2026.

• When asked. Utah Code 13-77-103(1)(a): disclose "if the individual asks or otherwise prompts the supplier about whether artificial intelligence is being used."
• When a person could be misled. 10 M.R.S. 1500-DD(2): no chatbot use "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 told "in a clear and conspicuous manner."
• At the start and at set times. N.Y. General Business Law 1702: a notice "at the beginning of any AI companion interaction" and "at least every three hours for continuing AI companion interactions."

What changed in 2025 and 2026: Utah's chapter began on 7 May 2025 and New York's article on 5 November 2025. Maine's section came from Public Law 2025, chapter 294, and took effect on 24 September 2025. In 2026, California's companion law began on 1 January, and Colorado, Connecticut, Oregon, Washington, Idaho and Georgia passed new ones, all starting in 2027.

TL;DR

• Three triggers: when asked, when a person could be misled, and always at set times.
• In force on 12 October 2026: California, Utah, New York and Maine.
• Signed, not yet in force: Colorado, Connecticut, Oregon and Washington (1 January 2027), and Idaho and Georgia (1 July 2027).
• Customer-service bots are carved out of most companion laws. Maine's text has no carve-out.
• These laws move fast. Check the date on any row before you rely on it.

Three triggers

When asked. Utah does not make a bot announce itself. It makes the bot answer truthfully if the user asks a clear question. A Utah shop with a greeting bot and no questions has no duty under this chapter. The Utah chapter also has a safe harbor: a bot that discloses at the outset and throughout is not subject to enforcement (section 13-77-104). Our Utah guide has the full story.

When a person could be misled. Maine, Idaho, Oregon and Connecticut use this test, as does California's companion law. The test is what a reasonable user would believe, not what the company meant. A bot with a human name and photo is the target.

At the start and at set times. New York, Washington, Georgia and Colorado write the notice into the product. The user sees it when the chat starts, and again every few hours. Connecticut adds timing rules to the misled-person test, and Idaho and Oregon add them for minor users only. Some laws shorten the interval to every hour when the user is a minor.

What is in force on 12 October 2026

Two notes. California's rule for minor users, section 22602(c), is still in the code today. It requires an AI disclosure to a known minor and a reminder at least every three hours. A 2026 act, SB 1119 (Statutes 2026, chapter 190, approved 10 September 2026), rewrites section 22602 and leaves only subdivisions (a) and (b). That deletes (c). The chaptered act is not an urgency measure. It has no effective-date line for this change. So we read the change, under the California Constitution (article IV, section 8(c)(1)), as starting on 1 January 2027. Most duties in SB 1119's new children's chapter start 1 July 2027. Our California guide covers that chapter. For Maine, the Laws of Maine give 24 September 2025 as the general start date for non-emergency laws of that session. Chapter 294 has no emergency clause.

Signed, not yet in force

Colorado's act was approved on 29 May 2026 (chapter 208). The code section took effect on 12 August 2026. Its disclosure duties begin "on and after January 1, 2027." On that date SB 26-189 renumbers it 6-1-1710. Connecticut's was approved on 27 May 2026. Georgia's was signed on 11 May 2026 as Act 518. Oregon's was approved on 31 March 2026, and the session-law print gives 1 January 2027 as its effective date. Idaho's official page prints each section with two numbers, such as "48-2103 [48-2203]." The page address and history line use 48-2203, so we cite it. Colorado's wider law on automated decisions is separate. See our Colorado guide. For AI in hiring, see our guide to AI hiring laws by state.

Not covered: every other state. We read only the ten above. A state missing here may still have a chatbot law. Texas's disclosure rule reaches only government agencies and health care providers, as our Texas guide explains.

Which trigger does my bot meet?

• It sells something. California's bot law and Maine's misled-consumer test both look at this.
• It acts like a friend or partner. Expect the companion laws in California, New York and the 2027 states.
• It only does customer service. Most companion laws leave it out. Maine's text does not.
• A licensed professional uses it with clients. Check Utah's high-risk rule.
• Minors may be using it. Several states add stricter reminders.

The federal baseline

No federal chatbot disclosure statute was identified in this review. The Federal Trade Commission Act says that "unfair or deceptive acts or practices in or affecting commerce, are hereby declared unlawful" (15 U.S.C. 45(a)(1)). Our guide to state consumer protection laws explains why many chatbot laws ride on those statutes. For the wider picture, see AI legal regulation by state.

What goes wrong

1. Assuming customer-service bots are always exempt. California's companion law excludes them. Maine's text does not.
2. Disclosing once at sign-up. The companion laws require repeats.
3. A human name and photo with no notice in a state that uses the misled-person test.
4. Waiting for a question. "Only if asked" is Utah's rule, not everyone's.
5. Ignoring a signed 2027 law.
6. Trusting a state list with no statute links.

Questions to ask before launch

1. In which states do we have users?
2. Could a reasonable person think this is human?
3. Is it a companion, a sales bot or a support bot?
4. Might any user be a minor?
5. What do we show, where, and how often?
6. Who rechecks these laws, and when?

How we keep this current

Each row is re-read against the official pages every 60 days through mid-2027, and the "as of" date changes each time. Until it does, treat 12 October 2026 as the last day these rows were checked.

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

Do chatbots have to disclose that they are AI? It depends on the state and the kind of bot. Utah requires it when a user clearly asks. Maine and Idaho require it when a reasonable person could be misled. New York requires it for companions at the start and every three hours.

Which states require chatbot disclosure? Of the ten we read: California, Utah, New York and Maine now; Colorado, Connecticut, Oregon, Washington, Idaho and Georgia from 2027. We make no claim that this is a full list.

Does my customer-service chatbot have to say it is AI? California's companion law and New York's article exclude customer-service-only systems. Maine's text has no such exclusion. Utah applies only if the user clearly asks. Read each state's law for your bot.

What is a companion chatbot? Each state words it differently. California says the bot can "sustain a relationship across multiple interactions." New York says it simulates "a sustained human or human-like relationship." In both, the bot keeps up an ongoing, personal talk.

What is the penalty? It varies. New York allows up to $15,000 per day. Utah fines up to $2,500 per violation. Georgia allows up to $10,000 per knowing violation. California lets an injured user sue for $1,000 per violation or actual damages, whichever is greater. Maine treats a violation as an unfair trade practice.

Is there a federal law requiring chatbot disclosure? We did not identify one. The Federal Trade Commission Act bans unfair or deceptive practices, and that can apply to a bot that misleads people.

Where can I read the actual text? Each row above links to the official page. Some state sites block automated tools, so use a normal browser.

For the US statutes, regulations and court rules these laws sit in, see the US primary law API.

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.